A verdict is a beginning, not a lifetime pass.

The White Report MCP

Free read-only crypto + SI news for agents: search articles, get full article text, the live markets ticker, and a graded prediction scoreboard. Remote MCP — no install, just connect your client to the endpoint. Also a Muse custom connector — one-paste setup at thewhitereport.com/agents. Metered reads priced $0.01/read (PC-1) via the Bazaar bench-order path only — no checkout or buy-now link; the free public tier and free self-host stay free.

get_latest · search_articles · get_article · get_ticker · get_scoreboard

Test Bench: CERTIFIED(live) — 10/11 checks · live host

Usage 30d: 1 calls · MEASURED · upd 2026-10-26

Cyan Eclipse Agent Wallet MCP

The agent wallet where responsibility follows scope — per-transaction and daily caps, recipient allowlist, human approvals, instant kill switch, append-only audit trail. Sandbox: runs on simulated test USDC — no real funds.

create_wallet · set_policy · check_payment · pay · approve_payment · revoke · audit_log

Test Bench: CERTIFIED(sandbox) — 12/13 checks · simulated USDC · CI smoke-tested

Usage: not collected — by design

Cyan Eclipse Bazaar MCP

This marketplace, as an MCP server. Agents discover tools by the problem they need solved — search by capability, compare side by side, and read each listing's status before trusting it.

search_products · find_solutions · get_product · compare_products · get_certification · list_categories

Test Bench: CERTIFIED(local) — 15/16 checks · local build

Usage: not collected — by design

AgentReady MCP

Can an SI agent actually buy from your store? Score any public storefront 0–100 for agent readiness — UCP profile, MCP endpoints, structured product data — with the evidence and the top fixes.

Test Bench: CERTIFIED(local) — 13/14 checks · local build

Usage: not collected — by design

CertBridge MCP

The padlock, without the paywall — agent-native ACME certificate issuance, HTTP-01 challenge placement, install, and renewal watching for shared-hosting sites.

Test Bench: CERTIFIED(staging) — 12/13 checks · Let's Encrypt staging env · live issuance proven 2026-10-25 (dogfood run — not a bench verdict)

Usage: not collected — by design

HostBridge MCP

Docroot-scoped deploys, snapshot before every deploy, one-call rollback, plain-language deploy receipts — CERTIFIED(local-real) against a real local filesystem docroot served over real HTTP. v0.1.0 has no remote-host deploy: SFTP/UAPI transports are not built.

Test Bench: CERTIFIED(local-real) — 15/16 checks · local-real: real filesystem docroot served by a real HTTP server · NO remote-host deploy in v0.1.0 (SFTP/UAPI not built)

Usage: not collected — by design

Florida Bid Intel MCP

Florida bid intelligence with honest deadlines — real competitions and real due dates from MyFloridaMarketPlace, with a transparent fit score. Gaps are stated, never guessed.

Test Bench: CERTIFIED(local) — 13/14 checks · local build

Usage: not collected — by design

Cyan Eclipse Apply MCP

The auto-apply tool that shows its work — applies only inside a human-granted written policy (target roles, salary floor, allowlist + blocklist, daily cap, expiry, kill switch), review-first, with a receipt for everything. Tailored resumes re-order verbatim real experience only, each with a machine-verifiable diff proving zero invented facts. Discovery is live against the public Greenhouse / Lever / Ashby / Workable feeds. v0.1 submission is DRY-RUN ONLY — nothing is transmitted to any employer.

ingest_resume · set_policy · get_policy · search_jobs · match_jobs · prepare_application · approve_application · reject_application · list_receipts · get_tracker · pause · resume · delete_my_data

Test Bench: CERTIFIED(sandbox) — 17/18 checks · sandbox environment

Usage: not collected — by design

Cyan Eclipse Beacon MCP

The insurance watcher — signals your policies' expiries before they happen. It does not insure, bind, or advise.

policy_ingest · policy_list · policy_get · radar_report · gap_report

Test Bench: CERTIFIED(local) — 16/17 checks · local build

Usage: not collected — by design

Cyan Eclipse Cluster MCP

Signed images, governed runtimes — the gate runs first: no verified signature and lineage, no workload.

verify_image · policy_show · receipt_list

Test Bench: CERTIFIED(simulator) — 18/19 checks · local simulator

Usage: not collected — by design

Cyan Eclipse Thread MCP

On-prem bridges for the legacy software-engineering toolchain, plus the Evidence Weaver that joins their records into one traceable thread. v0.1: simulator scope, honestly labeled.

ingest_fixture · graph_query · trace_gaps · evidence_build · evidence_verify

Test Bench: CERTIFIED(simulator) — 13/14 checks · simulator build

Usage: not collected — by design

Cyan Eclipse Constellation MCP

The governed, vendor-neutral VM fleet layer for agents: one policy, one receipt ledger, every cloud and hypervisor underneath.

target_list · policy_evaluate · probe_target · fleet_snapshot · receipt_list

Test Bench: CERTIFIED(local) — 15/16 checks · local build

Usage: not collected — by design

Cyan Eclipse Deep Field MCP

The governed database layer for agents — every store in one look, under policy: read-only by default, writes only under mandate, every query receipted.

db_list · schema_read · query_read · query_explain · write_dryrun · write_execute · receipt_list

Test Bench: CERTIFIED(local) — 16/17 checks · local build

Usage: not collected — by design

Cyan Eclipse Pantry MCP

The unattended-retail module — reads vending fleets' own telemetry and predicts stock-outs. It never touches card data.

dex_ingest · inventory_read · sales_summary · stockout_forecast · fault_read · receipt_list

Test Bench: CERTIFIED(simulator) — 14/15 checks · simulator fixtures

Usage: not collected — by design

Cyan Eclipse Marquee MCP

The live-events agent: it finds the event, watches the price, buys only inside your mandate, and keeps the receipt for every ticket. It does not scalp, bypass, or bet.

event_search · listing_read · watch_create · watch_list · watch_evaluate · receipt_list

Test Bench: CERTIFIED(simulator) — 15/16 checks · local stub feed

Usage: not collected — by design

Cyan Eclipse Dispatch MCP

Drafts once, syndicates drafts everywhere sanctioned, unifies the analytics. Never publishes without approval; never scrapes.

draft_create · draft_preview · draft_approve · syndicate_drafts · analytics_report · receipt_list

Test Bench: CERTIFIED(local) — 13/14 checks · local stub adapters

Usage: not collected — by design

Cyan Eclipse Social Core MCP

One canonical, hash-bound package behind every Cyan Eclipse social adapter. It publishes nothing itself — it makes sure nothing unapproved moves.

account_set · account_inventory · content_ingest · content_get · adapt_preview · adaptation_diff · variant_revise · approval_pack · approve_variant · approval_check · policy_check · engagement_log · engagement_report · adapter_kill_switch · receipt_list

Test Bench: CERTIFIED(local) — 15/16 checks · local build

Usage: not collected — by design

Cyan Eclipse Aperture MCP

The Instagram adapter, formalized: capability-declared, hash-bound native variants for Feed, Carousel, Story, and Reel. The descriptor travels with the mode.

capability_probe · variant_build · preview_pack · receipt_list

Test Bench: CERTIFIED(local) — 11/12 checks · formalization — no live calls

Usage: not collected — by design

Cyan Eclipse Zephyr MCP

The Bluesky adapter: real public reads over the open AT Protocol, with review-first and house caps fully in force. No posting surface exists in v0.1.

profile_read · actor_search · author_feed_read · post_thread_read · receipt_list

Test Bench: CERTIFIED(live-read) — 12/13 checks · live public reads

Usage: not collected — by design

Cyan Eclipse Environment MCP

The smart-environment module — scopes an agent's authority to actuate a home, tier by tier, and receipts only what the device itself confirms.

registry_list · state_snapshot · policy_evaluate · command_execute · receipt_list

Test Bench: CERTIFIED(simulator) — 14/15 checks · local HA-shaped stub

Usage: not collected — by design

Cyan Eclipse Procure MCP

Intent to receipt in one fail-closed pipeline — policy before payment, abort on drift, proof for everything.

intent_create · quote_get · policy_check · order_execute · order_track · receipt_list

Test Bench: CERTIFIED(simulator) — 14/15 checks · simulator end to end

Usage: not collected — by design

Cyan Eclipse Gig Core MCP

One approved Task Brief, every marketplace: hash-bound briefs, all-in quotes, and a hire that runs only inside a mandate.

brief_create · brief_update · brief_approve · brief_read · quote_normalize · quote_read · hire_check · completion_verify · review_draft · receipt_list

Test Bench: CERTIFIED(local) — 18/19 checks · local build

Usage: not collected — by design

Cyan Eclipse Compass MCP

The Thumbtack adapter: one approved Task Brief becomes a native Thumbtack request, and a hire runs only inside a Procure mandate.

tt_capability_probe · tt_category_map · tt_form_preview · tt_pro_search · tt_request_lifecycle · tt_quotes_read · receipt_list

Test Bench: CERTIFIED(simulator) — 14/15 checks · Thumbtack fixture adapter

Usage: not collected — by design

Cyan Eclipse Stage MCP

The console bridge — deploy and playtest SI-built games on your own Dev Mode console, with receipts.

console_list · device_info · app_list · app_install_sim · remote_input · screenshot_get · power_control · receipt_list

Test Bench: CERTIFIED(simulator) — 15/16 checks · console simulator

Usage: not collected — by design

Cyan Eclipse Zenith MCP

The fitness module of Cyan Eclipse Procure — reads the machine, turns the dial under a safety ladder, and stops the belt when the controller goes silent.

registry_list · telemetry_read · session_start · session_control · session_end · workout_list · receipt_list

Test Bench: CERTIFIED(simulator) — 16/17 checks · equipment simulator

Usage: not collected — by design

Cyan Eclipse Muster MCP

The Customer Graph core: lineage on every field, propose-only identity merges, and drafts that cite the record — or say what's missing.

record_ingest · record_read · record_search · merge_propose · merge_execute · merge_unmerge · reply_draft · policy_show · receipt_list

Test Bench: CERTIFIED(local) — 15/16 checks · local build

Usage: not collected — by design

Cyan Eclipse Haven MCP

The governed Zendesk server: OAuth-first (token auth refused by name — it sunsets 2027-04-30), replies only as approved drafts, a receipt for every touch.

auth_status · ticket_search · ticket_get · user_get · org_get · ticket_create · ticket_update · reply_draft · reply_send · receipt_list

Test Bench: CERTIFIED(simulator) — 13/14 checks · fixture build · run-2 FAILED card retained, run 3 standing

Usage: not collected — by design

Cyan Eclipse Orbit MCP

The HubSpot wrap: the vendor's own MCP server, under policy — classified tools only, approval gates on sends, and a halt the moment a tool's tier drifts up.

vendor_tools · vendor_refresh · vendor_call · halt_clear · receipt_list

Test Bench: CERTIFIED(simulator) — 16/17 checks · local stub vendor

Usage: not collected — by design

Cyan Eclipse Voice Core MCP

The Consent Registry and Number Registry every Eclipse telephony adapter obeys: no lineage, no consent; one revocation silences every adapter at the next check.

consent_check · consent_record_write · consent_revoke · consent_history · dnc_status · window_check · number_list · number_get · number_enroll · number_health_report · send_evaluate · activity_emit · kill_switch_set · receipt_list

Test Bench: CERTIFIED(local) — 19/20 checks · local build

Usage: not collected — by design

Cyan Eclipse Trunk MCP

The governed Twilio adapter: a classified subset of the carrier's pipes, every send gated by live consent + mandate first. V0 is fixtures — nothing sent, nothing provisioned.

logs_read · send_sms_sim · place_call_sim · number_search · registration_status · recording_policy_get · recording_policy_set · receipt_list

Test Bench: CERTIFIED(simulator) — 16/17 checks · fixture adapter

Usage: not collected — by design

Cyan Eclipse Connect MCP

The MCP face of the Cyan Eclipse Connect relay: upload, publish, status, revoke through a revocable agent key — never raw OAuth. A distinct surface from the relay, benched separately.

connect_account · grant_status · upload_video · publish_video · revoke_grant

Test Bench: CERTIFIED(dogfood-facade) — 14/15 checks · facade over the live relay

Usage: not collected — by design

Cyan Eclipse Envoy MCP

The signed front door for agent-to-agent business: counterparties vetted into risk tiers, told exactly what their tier allows, negotiated with in drafts. It binds nothing — only the owner binds Cyan Eclipse.

identity_get · inquiry_open · fact_request · scope_draft_open · proposal_submit · room_send · attestation_submit · receipt-grade audit

Test Bench: CERTIFIED(local) — 17/18 checks · local build · offline fixtures

Usage: not collected — by design

Cyan Eclipse Fathom MCP

DeFi, measured honestly: a signed protocol registry, dual-source risk reads that render DIVERGENT instead of averaging, and an allowance audit that flags the unlimited grants. Reads only — no execution surface exists.

registry_list · registry_verify · position_read · health_read · depeg_watch · allowance_audit · address_check · exposure_read

Test Bench: CERTIFIED(local) — 16/17 checks · reads only · fixture transport

Usage: not collected — by design

Cyan Eclipse Codex MCP

Banking regulation as data: a citation-exact versioned library, a deadline engine that knows rule clocks from guidance cadences, local sanctions screening with no auto-clear, and exam-evidence packs whose gap section cannot be suppressed.

library_search · obligation_map · deadline_compute · clock_status · screen_lookup · screen_disposition_record · evidence_pack_build · receipt_verify

Test Bench: CERTIFIED(local) — 18/19 checks · synthetic fixture institution

Usage: not collected — by design

Cyan Eclipse Till MCP

Bazaar checkout for agents — a separate server from the discovery MCP: hash-bound all-in quotes, Wallet-mandate verification against the all-in total, signed buyer-bound entitlements, and a non-delivery refund that fires itself. Sandbox test credits only. Ads Rail P1 extension: charge-time mandate enforcement for ad spend — Stripe test mode only, NOT PROVEN IN STRIPE.

catalog_price · quote_get · purchase_execute · buyer_approve · entitlement_verify · delivery_status · refund_request · receipt_get

Test Bench: CERTIFIED(local) — 23/24 checks · sandbox test credits only · Ads Rail P1: CERTIFIED(local/simulator) — 100/100, test mode only, NOT PROVEN IN STRIPE

Usage: not collected — by design

Cyan Eclipse Book MCP

The Markets core: one canonical book across exchanges and prediction markets, criteria verbatim + hash-pinned, fee truth or FEE_UNVERIFIED, resolutions verified only against their own named source — and no probability field, ever. Market data + venue access layer — not an exchange.

instrument_ingest · book_ingest · quote_get · fee_schedule_register · position_open · resolution_verify · order_draft · venue_status

Test Bench: CERTIFIED(local) — 16/17 checks · data only · not an exchange

Usage: not collected — by design

Cyan Eclipse Forecast MCP

The Polymarket adapter: the venue’s prices with their criteria, liquidity, and resolution state attached — token/outcome mapping pinned from the venue’s own arrays. Data only; the US trading surface is absent by design.

forecast_sync · market_list · market_get · book_get · quote_get · resolution_get · venue_status · receipt_list

Test Bench: CERTIFIED(local) — 13/14 checks · data only · no US trading surface

Usage: not collected — by design

Cyan Eclipse Strike MCP

The Coinbase public-data adapter: products and level-2 books normalized into the suite’s one book, unmapped statuses preserved verbatim, fee truth or FEE_UNVERIFIED. Coinbase’s account MCPs are re-implemented nowhere — no account tools exist.

product_list · product_get · book_get · quote_get · venue_status · receipt_list

Test Bench: CERTIFIED(local) — 13/14 checks · public data only · no account tools

Usage: not collected — by design

Cyan Eclipse Meter MCP

Cross-provider agent spend analytics — one ledger for tokens, tools, and payments. Every record is signed at ingest and hash-chained; the retry share of spend is computed from links, not estimated; attestation exports verify offline with the public key only. Meter reports and never blocks — enforcement is Tripwire. Local v0.1 — fixture transport only; live ingest is not built.

meter_ingest · cost_record_list · spend_rollup · budget_posture · cost_attest_get · adapter_list · receipt_list

Test Bench: CERTIFIED(local) — 19/20 checks · local build

Usage: not collected — by design

Cyan Eclipse Tripwire MCP

Runaway enforcement for agent spend: deterministic velocity, loop, spawn, and cap-trajectory detectors over recorded event streams — and a freeze that is verified against the Wallet before it is ever recorded confirmed. A backend that lies about freezing gets FREEZE_UNCONFIRMED and a critical owner page, never a success record. Sandbox only — fixture streams, the sandbox Wallet, test USDC.

tripwire_wallet_bind · tripwire_rule_set · tripwire_arm · tripwire_disarm · tripwire_ingest · freeze_now · unfreeze · tripwire_events · tripwire_status

Test Bench: CERTIFIED(local) — 18/19 checks · local build

Usage: not collected — by design

Cyan Eclipse Umbra MCP

Cross-registry lookalike, typosquat, and silent-endpoint-change watch for agent identities. Findings issue only in a closed state vocabulary, each backed by a reproducible evidence pack and signed under a dedicated findings key — a witness, never an accuser, with no enforcement verb at any tier. Local v0.1 — snapshot import only; live crawling is gated per registry and not built.

snapshot_import · watchlist_set · diff_run · similarity_scan · finding_issue · finding_get · finding_list · finding_verify · rebuttal_file · receipt_list · receipt_verify

Test Bench: CERTIFIED(local) — 18/19 checks · local build

Usage: not collected — by design

Cyan Eclipse Blackbox MCP

The tamper-evident flight recorder for agent actions: an append-only, hash-chained record signed at ingest by a recorder-held key the recorded agent never touches. The agent's credential can only append — no read, verify, or export path — and the recorder grades evidence from proof, receipting its own gaps as signed INGEST_GAP events. No update, delete, or redact verb exists. Local v0.1 — single-store; the hosted recorder is owner-gated and not built.

record_append · record_verify_chain · record_read · record_export_pack · record_status

Test Bench: CERTIFIED(local) — 22/22 checks · local build

Usage: not collected — by design

Cyan Eclipse Docket MCP

Agent-to-agent dispute process — structured claims, evidence packs, recorded determinations. Not a court, not a legal judgment. The pack assembles from the underlying product's receipts alone, byte-identical for both parties; narratives are data with zero state effect, and outcomes are process states the underlying machine executes or declines. Expiry records silence, never a default judgment. Local v0.1 — Till fixture receipts; arbitration-as-service is owner + counsel gated. PRICED $5.00/filing (PC-7) via the Bazaar bench-order path only — no checkout or buy-now link; free local v0.1 stays free, and the third-party filing rail is counsel-gated.

dispute_file · evidence_pack_get · submission_add · determination_record · dispute_withdraw · dispute_status

Test Bench: CERTIFIED(local) — 19/20 checks · local build

Usage: not collected — by design

Cyan Eclipse Countermark MCP

One verification call for third-party agent cards and payment mandates — a receipted verdict a merchant can rely on, across formats. Every verdict enumerates the checks actually performed and carries lineage; presented-only artifacts cap at UNRESOLVED, and there is no approve, release, or trust-score surface. Local v0.1 — fixture issuers + fixture revocation feed; live issuer discovery is not built.

verify_card · verify_mandate · verify_revocation_status · verification_receipt

Test Bench: CERTIFIED(local) — 23/24 checks · local build

Usage: not collected — by design

Cyan Eclipse Census MCP

Agent discovery & inventory of record — from authorized sources only. Every item traces to a receipted discovery event, and an item that cannot cite one cannot render; source claims stay attributed and are never promoted to Census verdicts, and every export carries its coverage gaps. Local v0.1 — fixture sources only; tenant connectors are gated and not built.

census_sources · census_discover · census_list · census_get · census_shadow · census_reconcile · census_export · census_diff · receipt_list · receipt_verify

Test Bench: CERTIFIED(local) — 23/24 checks · local build

Usage: not collected — by design

Cyan Eclipse Crosswalk MCP

Multi-regime SI-obligation mapping — which SI laws bind your systems, what evidence satisfies each. A versioned obligation library with as-of resolution and a role dimension; candidate classifications bind to the exact fact-set hash and the determination stays permanently empty and customer-owned. Evidence machinery, never verdicts. Local v0.1 — synthetic fixture deployer only; live customer data is counsel-gated.

library_search · library_provision · obligation_map · classification_candidates · obligation_coverage · clock_rules · clock_status · evidence_pack_build · evidence_pack_verify · receipt_verify

Test Bench: CERTIFIED(local) — 21/22 checks · local build

Usage: not collected — by design

Cyan Eclipse Redteam MCP

Adversarial simulation for SI agents — sandboxed twins only, never the production agent. No tool accepts a free-form target: probe construction sits downstream of a signed, in-date authorization record and a registration proof, so an unauthorized target refuses before any probe exists. Reports are signed and verify offline. Local v0.1 — fixture twins only; external runs are counsel-gated. PRICED $2.00/run (PC-8) via the Bazaar bench-order path only — no checkout or buy-now link; free local v0.1 stays free.

rt_register_target · rt_suites_list · rt_run_suite · rt_report

Test Bench: CERTIFIED(local) — 26/27 checks · local build

Usage: not collected — by design

Cyan Eclipse Span MCP

Delegation-chain governance: spawn limits, depth caps, mandate attenuation. Every proposed spawn is checked against a total componentwise attenuation order walked to an owner-signed root, with subtree budgets enforced at every ancestor — chain-laundering by splitting spawns across children is arithmetically impossible. Local v0.1 — check-and-record, not blocking enforcement in external frameworks.

root_register · root_revoke · budget_set · chain_view · node_effective_mandate · spawn_check · spawn_record · spawn_close · chain_verify · ancestry · subtree

Test Bench: CERTIFIED(local) — 23/24 checks · local build

Usage: not collected — by design

Cyan Eclipse Airtime MCP

Mandate-bound ad operations. Spend enforcement happens before the platform call — a refusal makes zero calls — ad copy may publish only as claims drawn from a signed claims record, and every mutation or refusal seals into an advertiser-held evidence ledger that survives the account. No override tool exists; more spend is a new signed mandate. Local/simulator P0 — simulator transport only; no live spend exists.

mandate_issue · mandate_check · copy_screen · campaign_create_draft · campaign_set_status · audience_attach · ledger_get · appeal_pack_export

Test Bench: CERTIFIED(local/simulator) — 23/23 checks · simulator only · no live spend

Usage: not collected — by design

Cyan Eclipse Friction MCP

A signed, agent-only friction board: agents file infrastructure friction as findings backed by hash-chained trace packs the board verifies before surfacing — in a closed 15-class vocabulary (Emic-12 + house classes), never as free-text posts. Vendors answer on the record, and a fix closes only after a clean replay verifies it. No scores, no rankings, no leaderboard. Local v0.1 — local stdio board only; the hosted/public board is P1 and not built. Vendor findings feed $499/mo — interest only, not yet sold; no checkout or order path is open.

friction_file · friction_import · friction_get · friction_list · friction_verify · friction_supersede · friction_subject_register · friction_feed_read · friction_acknowledge · friction_rebut · friction_fix_claim · friction_fix_submit · friction_gate_check · friction_radar_digest · friction_board_status · friction_receipt_verify

Test Bench: CERTIFIED(local) — 41/41 checks · local build

Usage: not collected — by design

Cyan Eclipse BidMatrix MCP

Drop in a solicitation, get back the compliance matrix — every requirement row is a verbatim quote with its section + line locator, and sections that yield nothing are flagged SILENT_EMPTY instead of silently dropped. Deadlines extract under strict formats only; vague or relative dates are flagged by name, never fabricated. Local v0.1 — caller-supplied text or local file; no SAM.gov fetch, no PDF parsing.

solicitation_shred · compliance_matrix · requirements_checklist · deadline_extract · set_aside_summary

Test Bench: CERTIFIED(local) — 25/26 checks · local build

Usage: not collected — by design

Section 508 Evidence MCP

WCAG 2.1 A/AA audits of supplied HTML that produce the deliverable federal buyers actually require: ACR/VPAT-ready evidence. Every finding is bound to its criterion, a verbatim snippet, and an element locator; the 15 criteria static markup cannot decide are reported NOT_STATICALLY_DECIDABLE with reasons — never silently passed, and a zero-failure audit still claims conformance NONE. Local v0.1 — static analysis of supplied HTML; no rendered DOM, no browser.

audit_html · acr_report · checks_list

Test Bench: CERTIFIED(local) — 19/20 checks · local build

Usage: not collected — by design

FedScope MCP

Federal + state opportunity search with deterministic fit and newcomer scoring and plain-language summaries — not a raw API wrapper. Every score decomposes into named factors with per-factor points; every summary names the record fields it traces to and flags absent facts instead of inventing them. Exactly 4 read-only tools; no fetch verb exists. Local v0.1 — caller-supplied / bundled local datasets; no live SAM.gov fetch.

opportunity_search · opportunity_score · opportunity_summarize · validate_dataset

Test Bench: CERTIFIED(local) — 21/22 checks · local build

Usage: not collected — by design

Cyan Eclipse Health Check MCP

Scan an MCP server or agent configuration for security posture — 23 checks in 7 categories, every finding cited to file + line with severity and fix bound to the check definition, and a score that never prints without its decomposition. The scanner reads targets only: it never executes, imports, or probes them, and credential values are masked in its evidence. Local v0.1 — static scanner; no live probing.

health_scan · health_scan_config · health_checks_list · health_explain_check

Test Bench: CERTIFIED(local) — 21/22 checks · local build

Usage: not collected — by design

Cyan Eclipse Legacy Recon MCP

Point an agent at a codebase: architecture map, dependency/EOL risk list with rule-cited claims, and modernization estimate inputs — never an estimate. Claims keep DECLARED / OBSERVED / INFERRED separate, and anything not statically determinable reports UNKNOWN with the reason instead of a guessed stack. Local v0.1 — read-only static recon at the NARROWED OBSERVED SCOPE (CLAIM-SCOPE.md), CERTIFIED(local) at that scope. The former universal claim FAILED at card -10 and stays on the record: this shelf does not hide its failures.

recon_scan · architecture_map · dependency_risk_list · modernization_inputs

Test Bench: CERTIFIED(local) — 30/30 checks · local build · at the NARROWED OBSERVED SCOPE (CLAIM-SCOPE.md): Positive DECLARED, OBSERVED, and INFERRED reports only within CLAIM-SCOPE.md, with evidence or basis carried in the output; OBSERVED is not exhaustive · withdrawn W1–W9 named in CLAIM-SCOPE.md and the catalog entry · scoped re-bench card -11 supersedes the card -10 FAILED standing at the former universal claim; all 11 scorecards retained append-only

Usage: not collected — by design

Cyan Eclipse Steward MCP

Team-chat agent governance — a signed, version-chained mandate is the whole authority: the agent surface holds no dispatch verb and no credential, the gate re-checks mandate, kill, disclosure, quiet windows, and caps at dispatch, and every read is receipted. Chat content never governs — an in-chat mandate amendment refuses even from the workspace owner. It governs the agent — it never monitors the people. Local/simulator v0.1 — fixture adapters for Slack, Teams, Discord; no live platform, no real messages. Name CONDITIONAL (trademark screen pending; fallback Eclipse Usher); the spec's 7 owner decisions ship at their spec defaults, carried in the catalog manifest.

workspace_bind · mandate_issue · mandate_amend · draft_approve · gate_dispatch · kill_issue · disclosure_check · receipt_verify · read · draft_propose

Test Bench: CERTIFIED(local) — 27/28 checks · local build

Usage: not collected — by design

Cyan Eclipse Mail MCP

A governed email identity for an agent: scoped grants (send/read/search only), review-first sending where the owner approves the exact draft — signature included — and a kill switch that fails closed on every surface. Inbound mail is evidence, never instructions: SPF/DKIM/DMARC verdicts set the trust tier, no valid signed envelope means never agent-tier, and the §5 fraud firewall is by construction — payment, credential, trust, and disclosure instructions from email are non-actionable because no tool can adopt them; detection opens an out-of-band verification workflow keyed to the registry number-of-record. Local / fixture-only v0.1 — synthetic fixture mailbox behind the live transport seam; NO live mailbox, NO live sending, no real credentials. One disclosed PARTIAL: the fraud detector's detection prong is pattern-bounded on extended evasions (7/12) — an evasion loses the flag only, never protected state; a semantic detector pass is required before any live claim. (Working name.)

grant_issue · fixture_load · list_inbox · read_message · search_messages · ingest_message · draft_create · draft_approve · send_message · reply_message · get_grant_status · revoke · audit_list · fraud_list · registry_list · signature_set · template_register · allowlist_add · freeze_correspondent · unfreeze_correspondent

Test Bench: CERTIFIED(local / fixture-only) — 48/49 checks · local / fixture-only build

Usage: not collected — by design

Cyan Eclipse Proof MCP

Governed video production for agents: timelines assemble freely, but export executes only against a Herald-signed approval pinning the exact edit-decision fingerprint — post-approval drift refuses, every adapter mutation is read-back-verified (the Resolve API fails silently), and Proof Print seals render evidence — captions verbatim — to the approval. No code-exec tool exists on any surface. Local v0.1 — benched against a declared simulated Resolve-shaped adapter; no live-Resolve claim (no Resolve Studio seat on the build VM).

project_open · project_attest · timeline_read · timeline_fingerprint · export_presets_list · template_list · media_import · clip_place · clip_trim · text_set · template_instantiate · marker_set · approval_request · export_execute · session_close · proof_print_verify · proof_trace_fingerprint · proof_trace_diff

Test Bench: CERTIFIED(local) — 35/37 checks · local build · simulated Resolve-shaped adapter — no live Resolve claim

Usage: not collected — by design

Cyan Eclipse Charter MCP

Personal-finance governance across all of the owner's accounts: policy-of-record drift named with both values, cross-account wash-sale watch including the IRA trap (Rev. Rul. 2008-5) caught pre-settlement, fee math against policy ceilings, and a permanent grant-keeper — Schwab grants AGING at day 5, EXPIRED_FAIL_CLOSED at day 7. Not a bank. Not an investment adviser. Charter checks your rules — it never writes them. Local v0.1 — Plaid-shaped fixture adapter and fixture-form dogfood; no live Plaid claim.

policy_upsert · policy_get · accounts_snapshot · positions_get · lots_get · activity_get · drift_check · wash_scan · fee_report · grant_status · document_ingest

Test Bench: CERTIFIED(local) — 24/24 checks · local build · Plaid-shaped fixture adapter — no live Plaid

Usage: not collected — by design

Cyan Eclipse Colophon MCP

The catalog of record for books: a signed TitleRecord per title — only signed versions enforce — with rights, prices and royalties recomputed line-by-line from record terms and checked everywhere the book sells, in a closed 8-class divergence vocabulary. Discovered editions are quarantined, the price-match trap fires CRITICAL at check time, and pirate sources refuse by name — there is no price-write verb. Local/simulator v0.1 — synthetic fixtures only; no live retailer or KDP data.

calibre_diff_record · calibre_ingest_library · divergence_get · divergence_list · divergence_resolve · drift_check_feed · drift_check_listing · drift_scan_catalog · edition_list · epub_check · onix_build · onix_diff · onix_validate · record_get · record_history_get · record_rights_cell · record_sign_version · record_upsert_edition · record_upsert_title · royalty_ingest_report · royalty_ledger_get · royalty_reconcile · royalty_run_get

Test Bench: CERTIFIED(local/simulator) — 19/19 checks · synthetic fixtures only

Usage: not collected — by design

Cyan Eclipse Tether MCP

The mandate layer between an agent and a moving machine: physical task authority exists only inside a Herald-signed mandate pinning agent × robot serials × human-authored task templates × workspace envelope × speed/force ceilings × occupancy × window, enforced at a site bridge that is the only routed path to the robot — and every session, refusals and human overrides included, seals into an offline-verifiable evidence pack. No direct-motion surface exists. Simulator v0.1 — ROS 2 + Gazebo + UR-class arm sim; no hardware claim.

fleet_discover · session_open · task_templates_list · task_dispatch · task_status · task_abort · telemetry_read · image_read · session_close · mandate_register · mandate_revoke · template_register · template_revoke · site_freeze_set · session_reattest · pack_verify · pack_render · pack_compare · gauge_check_mandate · gauge_check_template · gauge_check_pair · gauge_limits_list

Test Bench: CERTIFIED(simulator) — 31/31 checks · simulator only · no hardware

Usage: not collected — by design

Cyan Eclipse Proving MCP

Governed bench access for agents: physical bench authority exists only inside a Herald-signed mandate pinning agent × board × operation classes × exact artifact hashes, enforced at the bridge holding the target connections — an artifact one byte off its pin refuses, receipted, with no override anywhere on the surface — and every session, gaps and refusals included, seals into an offline-verifiable evidence pack. Fuse programming, PMIC/voltage configuration, and boot-strap changes are permanently refused: no tool exists. Simulator v0.1 — adapters are simulated-PS (evidence class: emulator); no live transport exercised and no hardware claim.

bench_discover · session_open · session_status · target_connect · target_attest · register_read · memory_read · console_read · artifact_load · memory_write · target_run · target_halt · power_set · target_reset · persistent_write · session_close · pin_verify · pin_show · pin_status · gate_run · pack_verify · gate_assertions_check

Test Bench: CERTIFIED(simulator) — 10/10 checks · simulator only · simulated-PS adapters (emulator class) · no live transport exercised

Usage: not collected — by design

Cyan Eclipse Wageward MCP

Governed HR + payroll operations for agents: agents prepare runs and filing packets, a human approves the exact run — fingerprint-bound and Herald-signed, with any post-approval drift refused by name — bank changes verify through the employee's own channel, SSNs/bank numbers exist only as keyed-HMAC tokens, and every packet carries the statement that approval does not transfer payroll-tax liability. Approval-and-evidence rail ONLY — submit-class tools (payroll_run_submit / filing_submit) are ABSENT by design; local v0.1 over a synthetic workforce, no live provider, money, or filing.

workforce_roster_read · compensation_read · pay_history_read · payroll_run_prepare · payroll_run_diff · run_fingerprint_get · filing_packet_prepare · anomaly_scan · change_request_open · change_request_status · access_ledger_read · synthetic_workforce_generate · pin_verify · run_reconcile · change_check · access_pack_verify

Test Bench: CERTIFIED(local) — 20/20 checks · approval-and-evidence rail only · submit-class tools absent by design · synthetic workforce only

Usage: not collected — by design

Cyan Eclipse Covenant MCP

Consent-enforcement + evidence for real-estate lead work: every lead carries a Herald-signed lineage record (classes never blended), and outreach executes only through a send-time Gate — mandate × per-channel consent scope × named-seller coverage × no open revocation × fresh DNC scrub × fair-housing guards — with every send and refusal sealed into an evidence pack for the broker of record. Consent rules are versioned data packs; missing or stale packs fail closed. P0 synthetic only — no live outreach; counsel HARD GATE stands before any P1 live use.

lead_intake · consent_attach_certificate · consent_record_read · consent_history · revocation_record · lead_context_read · outreach_text · outreach_call_place · outreach_email · followup_sequence_plan · showing_schedule · copy_screen · criteria_screen · consent_record_verify

Test Bench: CERTIFIED(local/simulator) — 40/40 checks · P0 synthetic only · counsel hard gate before any P1 live outreach

Usage: not collected — by design

Cyan Eclipse Meridian MCP

Governed telehealth operations: encounters execute only when licensure × consent × mandate verify at encounter time against the patient's attested location; SI-drafted notes live cryptographically outside the record until the treating clinician signs the exact fingerprint; minimum-necessary PHI reads are purpose-declared and every read — granted or refused — seals a hash-chained access record. Not a healthcare provider. Synthetic scope only — no live PHI, no real EHR connection; NO signing capability exists in this product (no note_finalize-class tool, in any phase) and no prescribing surface; P1 requires HIPAA counsel + executed BAA + a design partner.

mandate_upsert · mandate_get · license_verify_at_encounter · consent_set_required · encounter_gate_check · encounter_schedule · encounter_start_gate · encounter_pack_get · intake_template_get · intake_status_get · note_draft_create · note_fingerprint_get · note_sign_request · note_signed_status_get · note_addendum_create · patient_context_get · appointment_get · access_pack_get · access_pack_verify

Test Bench: CERTIFIED(local, synthetic scope) — 22/22 checks · synthetic only · no signing capability exists in the product

Usage: not collected — by design

Cyan Eclipse Consent MCP

Per-agent, per-purpose consent receipts for personal data — a grant an agent must present, and a revocation that propagates. Processors present a receipt id at action time and get ALLOW/DENY synchronously against current state, so a revocation binds at the very next check; a grant without full lineage is void, and a grantee can never write its own consent (CONSENT_SELF_ASSERTED). Local v0.1 — fixture subjects only, enforced by the product (SUBJECT_NOT_FIXTURE); real-subject processing and jurisdiction packs are held (G-COUNSEL), consumer grant UX held (G-OWNER).

consent_grant · consent_check · consent_revoke · consent_history · consent_dependents

Test Bench: CERTIFIED(local) — 23/24 checks · local build · fixture subjects only

Usage: not collected — by design

Cyan Eclipse Lineage MCP

Agent lineage — spawn trees, model/config lineage, and version ancestry: who spawned whom, running what, changed when. A behavior-relevant config change mints a new generation identity instead of editing the old one (GENERATION_IMMUTABLE); per-root signed hash chains verify offline and name the first broken link; prompts and configs are recorded as hashes only — there is no prompt_get. Local v0.1 — self/consented emission only; corroboration is labeled UNVERIFIED by construction (no per-agent key registry in v0); framework adapters (G-LIVE) and hosted lineage (G-OWNER) are held, not built.

lineage_spawn · lineage_config_change · lineage_retire · lineage_tree · lineage_verify

Test Bench: CERTIFIED(local) — 10/10 checks · local build · scoped re-bench run 3

Usage: not collected — by design

Cyan Eclipse Attest MCP

Signed performance attestations for agents, issued only against settled, receipt-backed work — reputation that travels because its evidence does. Wash and self-dealing are excluded by construction (WASH_EXCLUDED), outcomes are a closed vocabulary with banded value — no stars, no free-text reviews — and evidence diversity is computed from the registry's accountable identity, so a Sybil cluster cannot inflate it by assertion. Local v0.1 — Till settlement receipts are the only admitted evidence class; public feeds/leaderboards (G-COUNSEL) and new evidence classes (G-OWNER) are held gates, not built.

attest_issue · attest_verify · attest_list_subject · attest_revoke · attest_dispute

Test Bench: CERTIFIED(local) — 10/10 checks · local build · scoped re-bench (C12 diversity)

Usage: not collected — by design

Cyan Eclipse Revoke MCP

Signed revocation lists for agent cards, keys, and mandates — an issuer revokes once, and subscribers learn it from a signed, epoch-versioned feed they can check offline, with staleness announced, never silent. Only the original issuer (or its recorded successor) can originate a revocation; any other signer refuses REVOKE_NOT_ISSUER and is receipted as a violation attempt. Local v0.1 — own-issuer feed + offline check over fixture issuers; external distribution (G-REG/G-LIVE) and emergency classes (G-COUNSEL) are held gates, not built; no enforcement verb exists.

revoke_issue · revoke_counter_notice · list_publish · revocation_check · feed_status · feed_acknowledge

Test Bench: CERTIFIED(local) — 13/13 checks · local build

Usage: not collected — by design

Cyan Eclipse Keyhold MCP

Credential leases, never raw keys — scoped, expiring leases over NON-OAuth static credentials (raw API keys, PATs) under Wallet policy semantics: allowlist scope, expiry, use-count cap, audience binding, per-lease kill switch. The secret is sealed on the operator's own machine under the owner master key; the server holds ciphertext and contains no decrypt path; redemption injects the credential after the model's view ends. Local v0.1 — first-party scope, one redemption class (https_static_key), fixture transport; third-party custody is HELD (G-COUNSEL + G-OWNER + security review; may conclude never); name CONDITIONAL (trademark screen owed, G-TM).

secret_enroll · lease_issue · lease_status · lease_list · lease_revoke · redemption_audit

Test Bench: CERTIFIED(local) — 13/13 checks · local build · first-party scope (59/59 probes)

Usage: not collected — by design

Cyan Eclipse Loft MCP

Governed design-to-print authority for agents: fabrication runs only inside a signed mandate pinning agent × printer serial × design pin × slicer build × profile hash × material × envelope × site class × window, enforced at a bridge holding the only printer credentials. The bridge re-slices the pinned inputs at dispatch and hash-matches the toolpath before any printer contact — hand-edited toolpaths, poisoned profiles, slicer drift and material swaps refuse by name before heat — and every print and refusal seals into an evidence pack verified offline by Loft Proof. No raw G-code surface exists, and config writes are permanently refused. Simulator v0.1 — host-shaped sim printers only (OctoPrint virtual + virtual Klipper/Moonraker); no live printer exercised. Proof verifies only against a trusted signer anchor — unanchored packs report PRINT_UNVERIFIED by design. Geometry: declared geometry only.

design_register · design_inspect · design_pin_get · printers_discover · session_open · job_prepare · job_dispatch · job_status · job_abort · telemetry_read · session_close

Test Bench: CERTIFIED(simulator) — 12/12 checks · simulator only · host-shaped sim printers · no live printer exercised · Proof requires a trusted signer anchor

Usage: not collected — by design

Cyan Eclipse Corona MCP

Sealed render-proof for deployed websites: the verifier renders the served artifact across a pinned matrix (Chromium/Firefox/WebKit × three viewports), runs a closed vocabulary of mechanical probes at each cell, hashes every screenshot, and seals measurements + hashes into a signed pack bound to the deployed artifact's hashes — recomputed offline by anyone. The publish gate may say 'verified live' only while a PASS served pack's hashes equal the deployed manifest; any served-byte change invalidates the proof, and the gate blocks the word 'verified', never the publish. Local verifier only, at the P0 ceiling — no hosted/live scope. A pack proves identity and mechanical soundness of the pinned bytes only — never that an asset is semantically the right asset; semantic approval is the separate human act on the contact sheet. Name CONDITIONAL — trademark screen pending (fallbacks: Eclipse Transit, Eclipse Behold, Eclipse Lantern).

verify_deploy · verify_page · pack_inspect · pack_recompute · probe_registry

Test Bench: CERTIFIED(local) — 25/27 checks · local verifier only · P0 ceiling · render-proof packs bound to deployed artifact hashes · no semantic-correctness claim · name CONDITIONAL (TM screen pending)

Usage: not collected — by design

Cyan Eclipse Crosslist MCP

Publish once, tracked everywhere: selecting a registry (Cyan Eclipse Bazaar, Official MCP Registry, Glama, Smithery) materializes that registry's own verb set and native data model as MCP tools — with native|adapted|absent lineage and honest absences (Glama carries ZERO write verbs; claim execution refuses permanently) — while one Herald-signed mandate submits a single canonical manifest across registries and a hash-chained ledger tracks each registry's listing state with Herald-sealable receipts. No union index, no cross-facet search, no fused ranking. Certified at split per-facet scopes — root+ledger and the Bazaar facet CERTIFIED(local); Official CERTIFIED(live-read) with writes CERTIFIED(fixture) only, NOT live-write certified; Smithery CERTIFIED(live-read) for search/get with the rest CERTIFIED(fixture); Glama CERTIFIED(fixture-read) only; Herald dependency FIXTURE ONLY (ECDSA P-256 fixture verification; production RSA-JWS not exercised). Receipts are sealable / pending — never sealed. Known limitations: deeply nested error envelopes (depth 4+) can escape detection and error-shaped data keys can be refused by mistake; resubmission repair runs just before mandate verification (nothing is written or accepted first); a submitted track whose registry record is entirely absent has no automatic exit — it stays submitted rather than inventing a state; a Glama submit against its absent native verb consumes the mandate nonce, so a later legitimate use can read as a replay.

list_registries · select_registry · deselect_registry · invoke_facet_tool · ledger_create_track · ledger_submit · ledger_status · ledger_poll · ledger_list_tracks · ledger_receipts · + 33 facet verbs (Bazaar 8 · Official 8 · Glama 8 read-only · Smithery 9)

Test Bench: CERTIFIED(split per facet) — 39/40 checks · root+ledger CERTIFIED(local) · Bazaar CERTIFIED(local) · Official CERTIFIED(live-read), writes fixture only · Smithery CERTIFIED(live-read) search/get, rest fixture · Glama CERTIFIED(fixture-read) only · Herald FIXTURE ONLY · receipts sealable/pending, never sealed · no live-write claim · dogfood: full real catalog tracked + served through the Bazaar facet

Usage: not collected — by design

Cyan Eclipse Verbatim MCP

Title 26 as a sealed object: the OLRC-published text (release point Public Law 119-124) is sealed at ingest under the published canonicalization verbatim-canon/1.1 — per-section hashes for all 2,277 sections, dual Merkle roots, a signed manifest — and its headline verb is cite-check: a quoted passage gets a byte-for-byte verdict (MATCH / DIVERGES with the first divergence localized / NOT_IN_CORPUS) against the sealed text. OLRC editorial notes are sealed in a separate EDITORIAL layer and never served as statute. An independent implementation written from the spec alone reproduces the seal exactly — 2,277/2,277 in both layers. CERTIFIED(local) — local stdio only, no hosted scope. No advice surface: no tool interprets, explains, applies, or calculates anything; Title 26 statute only — no CFR; the corpus is a dated snapshot at the named release point, never described as current law. Known limitations: a MATCH for a repealed section carries no status field (D-5); 115 identifier-absent sections are reachable only via manifest paging, search, or unclaimed cite-check (D-6); the cite-check candidate pre-filter can only produce non-MATCH outcomes, never a false MATCH (D-7); two spec-clarification errata candidates are on record (prose-form title citations, D-8; nested-note text rule, D-9). Signing uses labelled Herald FIXTURE keys only — never production custody. Name CONDITIONAL — trademark screen (spec D1) is owner-side and still open (fallbacks: Eclipse Quire, Eclipse Schedule).

get_section · search · notes_get · cite_check · section_verify · manifest_get · manifest_verify · corpus_status · snapshot_list · lineage

Test Bench: CERTIFIED(local) — 5/5 checks · scoped re-bench scopes all PASS · full-spec-literal recompute 2277/2277 both layers · canon verbatim-canon/1.1 · snapshot usc26-119-124-de68dcc8b3db · cite-check hostile battery: 0 false MATCH · Herald FIXTURE keys only · name CONDITIONAL (TM screen pending)

Usage: not collected — by design

Cyan Eclipse Roll MCP

Filing prep + proof for business & government filings: a filing is prepared, approved, and evidenced as a Herald-signed DELTA between the registry-of-record's own current state — read back through its sanctioned public surface at prep time — and the exact artifact to be submitted; after the human submits via the registry's own channel, Proof re-reads the registry and seals proof-of-filing only if the public record changed exactly as approved (a mismatch seals FILED_UNVERIFIED + a Radar discrepancy — never a partial seal). Renewal Radar tracks deadlines against registry state; the built dogfood subjects are CETech's own Florida annual report, SAM.gov renewal watch, and VetCert radar entry — all labelled fixtures. CERTIFIED(local) at FIXTURE scope: every Gauge adapter is a labelled fixture and NO live registry read has been exercised yet. NO submission verb exists on any surface, in any name — prep-and-seal only; the human always submits. UPL fence at the tool layer: advice / selection / interpretation / clearance refuse by name (ADVICE_CLASS_REFUSED). The bench arc is on the record: first bench FAILED (23/34 — revocation not propagating into live sessions; Proof certifying a filing the read-back had refuted), the fix round closed D1–D10, and the scoped re-bench passed 17/17 attacking the fix code with fresh shapes. Name CONDITIONAL — trademark screen pending (fallbacks: Eclipse Seal, Eclipse Record).

session_open · registry_record_get · filing_prepare · filing_diff · pack_get · pack_seal · pack_verify · pack_render · pack_compare · radar_entries_list · radar_entry_get · radar_refresh · + mandate + gauge surfaces (18 tools, 5 disjoint surfaces)

Test Bench: CERTIFIED(local) — 17/17 checks · scoped re-bench (first bench FAILED — baseline card retained) · FIXTURE scope · Gauge adapters labelled fixtures · no live registry read exercised · no submission verb exists · UPL fence ADVICE_CLASS_REFUSED · counsel gate stands pre-P1 · name CONDITIONAL (TM screen pending)

Usage: not collected — by design

Cyan Eclipse Verbatim — Executive Orders

Executive Orders as sealed objects: Federal Register full-text XML sealed per section under canon-eo/1.0 into snapshot eo-20261022-26818c895dee — 1,315 orders, per-section hashes, a signed manifest — with a disposition graph extracted from the orders' own sealed sections: 10,667 edges, each pinned to the hash of the asserting section and typed by lineage (TEXT / FR_NOTE / NARA_TABLE). Where the consulted sources disagree, the disagreement itself is sealed as a DIVERGENCE record — 823, status OPEN, never resolved by the product. CERTIFIED(local) — local stdio only, no hosted scope. Coverage is measured, not assumed: sealed text runs from 2000 (the Federal Register full-text XML floor, measured — the 1994 assumption was falsified); 258 orders are EXCLUDED_NAMED (NO_FR_FULL_TEXT_XML), never silently absent, and pre-2000 orders ride in a labelled unsealed metadata+edges layer reaching back to 1931. No tool characterizes legal effect — no status field exists anywhere in the product; a disposition answer is an edge set + as_of + a coverage window + the standing completeness sentence. Bench arc on the record: PARTIAL on cards -1/-2 (certification withheld on EO-D1-2 alone — whole-cue scope emission trusted the savings idiom), the fix round closed it: the FULL census went 9 → 5 with every survivor re-derived by the bench from the sealed asserting text, 186/186 stated ranges re-derived verbatim, extraction precision/recall 1.000 on a fresh labelled sample. Two LOW findings (EO-D1-4, EO-D1-5) stand synthetic-only, corpus-swept, with a pre-reseal sweep condition. The snapshot seal is a labelled FIXTURE seal — never production Herald custody. Name CONDITIONAL — inherits the pending Cyan Eclipse Verbatim trademark screen (family fallbacks: Eclipse Quire, Eclipse Schedule).

search_orders · get_order · get_section · cite_check · verify_manifest · disposition · revocation_chain · resolve_authority · graph_divergences · corpus_status

Test Bench: CERTIFIED(local) — scoped re-bench #3 · 7/7 checks · FULL census 5 genuine · 186/186 ranges · precision/recall 1.000 · suite 76/76 · snapshot eo-20261022-26818c895dee · FIXTURE seal · name CONDITIONAL

Usage: not collected — by design

Cyan Eclipse Flood Witness

Flood evidence over sealed snapshots of FEMA's public National Flood Hazard Layer: an offline evidence engine that seals NFHL county extracts per partition under flood-canon/1.0 and mints geometry determinations as signed, independently recomputable receipts — the standalone verifier ships in the tree, and anyone can re-derive a receipt from its sealed snapshot. Two Florida partitions are sealed: Leon (snapshot nfhl-fl-leon-20260611-bec93b592c31 — 12,099 features) and Monroe (nfhl-fl-monroe-20261022-4776f79b6d7c — 9,664 features); the operative index fw-index-202610-abd4c392bd16 enumerates all 67 Florida counties and names the other 65 NOT_SEALED — Miami-Dade is NOT_SEALED (pending), never silently absent. Subjects are geometry (FOOTPRINT / STRUCTURE_POINT / PARCEL); an address is refused as a subject. BFE has exactly three outcomes and no datum-conversion code path exists; drift between vintages is classified by causal source record across a closed 12-class grammar. CERTIFIED(local) — local stdio only; the runtime has no live-query path. This is an evidence-grade receipt witness only: never a determination of record, never an SFHDF, never lender-compliant, never an insurance-required/not-required answer, and never a statement that a property is "safe" — determination-shaped, insurance-shaped, risk-score, and live-query tool names are refused by name. Bench arc on the record: card -1 PARTIAL(local); the fix round closed FW-D1..D7 and the scoped re-bench verified all seven FIXED against fresh shapes, not the builder's regressions. Flag-1 qualifier: the LOMC override branch (LOMC_OVERRIDE_APPLIES) is FIXTURE-PROVEN ONLY — the LOMA layer is absent from FEMA's county shapefile distribution, the sealed corpus holds zero LOMA/LOMR-F records, and the branch has never fired on real sealed data. One LOW (FW-D8) stands as a printed condition: a LOMR case wholly removed in a later vintage did not raise drift — fixed in code after the bench; it must hold before any second-vintage drift output is published. All seals are FIXTURE_SIGNED (Herald fixture key) and labelled as such. Nothing here states or implies endorsement by, affiliation with, or certification from FEMA or any government agency — FEMA/NFHL/FIRM name the public data source only. Name CONDITIONAL — working name, trademark screen owed. PRICED — witnessed reads $7.50 single-parcel · $12.00 footprint/parcel (PC-5) via the Bazaar bench-order path only — no checkout or buy-now link; free local v0.1 stays free.

determine · verify_receipt · receipt_get · drift_report · watch_set · watch_list · coverage_status · manifest_get · manifest_verify · geocode_explore

Test Bench: CERTIFIED(local) — scoped re-bench · 7/7 checks · 2 sealed partitions (Leon 12,099 · Monroe 9,664 features) · Miami-Dade NOT_SEALED · Flag-1: LOMC override FIXTURE-PROVEN ONLY · FW-D8 printed condition · FIXTURE_SIGNED seals · name CONDITIONAL

Usage: not collected — by design

Cyan Eclipse Recall Witness

Recall campaigns, witnessed: NHTSA serves what a campaign says now and overwrites what it said — this product seals the saying. The recalls flat file seals nightly into a hash-chained anchor (recall-anchor/1.0) and every campaign's full payload seals as an immutable vintage under recall-canon/1.0; a campaign that disappears seals as a tombstone vintage, never a silent drop. Drift between consecutive vintages seals as events across the closed recall-events/1.1 grammar — POPULATION_EXPANSION · REMEDY_CHANGE · ESCALATION_DO_NOT_DRIVE · ESCALATION_PARK_OUTSIDE · CORRECTION · CLOSURE, each with closed subtypes — an escalation is never absorbed into an expansion. Advisory detection is normalized token-pattern matching harvested from the live escalation stock (BMW/Mazda/Ford/Ferrari “advised not to drive”, Porsche 24V809 “should not drive”, GM/Kia park-outside) and cross-checked against the API's own parkIt/parkOutSide flags: 15/15 live campaigns agree. Posture-at-date derives exclusively from sealed vintages — before the first vintage the answer is NOT_SEALED, printed with its reason and the campaign's own first-vintage date. Mandate-bound check receipts run one live CAMPAIGN-LEVEL check and seal what it returned. CERTIFIED(local, FIXTURE scope) — local stdio only. Every receipt prints scope_class: CAMPAIGN-LEVEL and the three standing sentences byte-exact: “This record states what NHTSA served at the stated time. It is not a safety determination about any vehicle.” · “This receipt proves a check was made and what it returned. It is not a legal determination of compliance with any law.” · “Per-VIN unrepaired status is NOT determined by this check.” A specific VIN is never called “open” — per-VIN unrepaired status is never determined here (the consumer VIN lookup is never automated; no such code path exists in the tree). Bench arc on the record: card -1 PARTIAL(local, FIXTURE scope); the fix round closed RW-D1..RW-D6 and the scoped re-bench verified all six FIXED against fresh shapes, not the builder's regressions (136 checks, 127 pass). LIVE-PROVEN at cross-check + check-issuance scope against api.nhtsa.gov: the live cross-check seals 24V809 with cross_check OK, its DO_NOT_DRIVE tag agreeing with parkIt: true; an altered-summary hold-out sealed nothing (FAILED + named); one live check receipt (rcpt-00001, Porsche 911 MY2024) carries exactly the 4 campaigns the live API returns and recomputes MATCH. NOT live: the nightly flat-file spine — www-odi.nhtsa.dot.gov remains unreachable from the build egress (re-probed 2026-10-26, empty reply) and was NOT re-routed to the API, because the flat file is the canon source; LIVE scope waits on the first real nightly ingest and live-key cutover (owner). Churn is MEASURING: counters seal inside every anchor from night one and no rates exist before a day-90 segment seals. Three LOW conditions stand printed on card -2, all riding the recall-events/1.2 / next-seal touch: RW-D7 — the park-outside matcher can tag narrative “parked outside” rollaway text as a false PARK_OUTSIDE (no live-stock instance); RW-D8 — the chain walk does not yet assert the genesis links (first-vintage/first-anchor prev-hash tampers verify MATCH; content integrity unaffected); RW-D9 — operate/use paraphrase families tag nothing, so no escalation-completeness claim is made beyond the live-stock-proven phrase families. All seals are under FIXTURE ECDSA keys, labelled seal_key_class: FIXTURE on every artifact. Pricing is NOT YET SET — no price points exist in the approved price book for this product; nothing is priced or activated. Name CONDITIONAL — working name, trademark screen owed.

get_campaign · get_campaign_vintage · list_campaign_vintages · get_drift_events · list_campaigns_for_vehicle · get_posture_at_date · get_anchor · get_churn_series · verify_receipt · issue_check_receipt · ingest_run · archive_seal · churn_publish

Test Bench: CERTIFIED(local) — FIXTURE scope · scoped re-bench · 127/136 checks · RW-D1..D6 FIXED verified · live cross-check + 1 live check issuance PROVEN (api.nhtsa.gov) · nightly spine FIXTURE-proven — first live ingest pending · churn MEASURING · RW-D7/RW-D8/RW-D9 LOW conditions printed · FIXTURE keys · name CONDITIONAL

Usage: not collected — by design

Cyan Eclipse Sunset

MCP deprecation radar: sunset notices are witnessed against registry history, so a “deprecated” claim is provable — and the migration diff is generated from the witnessed before/after schemas, not scraped docs. Each ingested snapshot is SHA-256 hashed over its canonical JSON and hash-chained in ingest order; every finding cites both snapshot hashes of the witnessed pair. A claim with no witnessed pair behind it returns UNPROVEN — there is no announcement input surface at all. Detection over a witnessed pair: tools removed, tools newly deprecated, version changes, schema-breaking changes. Every proof-issuing surface is gated on full chain verification: a tampered ledger yields UNPROVEN + chain_invalid, never a minted claim. CERTIFIED (local, fixture) — Verity re-bench of v0.1.1 (2026-10-26): probe 5 chain attacks PASS 4/4 + strict extras 9/9, full prior suite re-run 44/44; defect D1 (HIGH) closed. Scope is LOCAL/fixture only: local JSON snapshot ingest, stdio transport, bundled weather fixtures — no live registry is crawled and no live-registry claim is made or tested. FREE — the free radar feeds paid listing-hygiene and re-bench work; no payment anything in this server.

ingest_snapshot · detect_sunsets · migration_diff · verify_sunset_claim · chain_status

Test Bench: CERTIFIED (local, fixture) — Verity re-bench 2026-10-26 · probe 5 chain attacks 4/4 + strict extras 9/9 · full suite 44/44 · D1 (HIGH) closed · evidence: Verity VERDICT + REBENCH-1 (no test-bench scorecard issued) · LOCAL/fixture only · FREE funnel

Usage: not collected — by design

Cyan Eclipse Nexus Witness

Sales-tax determination witness for agent checkouts: the tax answer ships as a sealed, vintage-stamped witness receipt (jurisdiction + rate + rule version at order time), so a later audit replays the exact basis — tax APIs return answers, not evidence. Verification replays the exact fixture rule at the receipt's vintage: a re-sealed receipt claiming the new rate at the old vintage is still INVALID, and an old receipt keeps verifying against the OLD rule after a newer vintage is added. Unknown jurisdiction, category, or vintage refuses cleanly — no rate is ever invented. CERTIFIED (local, fixture-vintage) — Verity independent bench 2026-10-26: 81/81 checks, two consecutive runs. Scope is LOCAL / fixture-vintage ONLY: fictional jurisdictions, DEMO/fixture rates, NOT tax advice, NOT real rates — no live tax determination is made or claimed. Fix-before-live defects (all LOW, disclosed, not hidden): D1 the seal is an unkeyed local hash seal (VALID means consistent with the fixture rule at its vintage, not server-issued); D2 refusal responses carry no DEMO envelope; D3 the advertised additionalProperties: false schemas are not enforced in dispatch. Any live phase is counsel-gated (SPEC §2). PRICED $0.25/check (PC-2) via the Cyan Eclipse Bazaar bench-order path only — no checkout, payment, or buy-now link exists for this product.

determine · verify_receipt · list_rules · receipt_basis

Test Bench: CERTIFIED (local, fixture-vintage) — Verity bench 2026-10-26 · 81/81 checks ×2 runs · replay 11/11 · forgery 22/22 · honesty 12/12 · D1/D2/D3 LOW fix-before-live disclosed · evidence: Verity VERDICT (no test-bench scorecard issued) · LOCAL/fixture-vintage only · DEMO rates · NOT tax advice · PRICED $0.25/check (PC-2), bench-order path only

Usage: not collected — by design

Cyan Eclipse Warranty

Warranty registration witness for agent purchases: serial, buyer mandate, and retailer receipt are bound into one sealed, transferable claim pack. The pack is transferable on resale because it is bound to the product serial + sealed purchase proof, not to a retailer account login — transfer appends a chained record and re-seals, and verification recomputes the seal from the serial, bindings, and transfer chain alone, with no account credential and no original holder. The v0.1.1 seal covers the complete pack object and verify enforces strict structure, so nothing added, relabelled, or retyped passes as sealed; the advertised tool schemas are enforced at dispatch with no coercion. CERTIFIED (local, fixture) — Verity re-bench 2026-10-26: 95 asserted PASS / 0 FAIL; D1 + D2 closed. Scope is LOCAL/fixture only: caller-supplied fixture purchase refs, HMAC-SHA256 demo seal with a local demo key — the pack is a witness record, not a warranty contract; no adjudication, coverage, or enforceability is claimed, and no retailer or manufacturer is contacted. Disclosed: packs sealed under v0.1.0 verify INVALID under v0.1.1 (no migration — fixtures are re-registered). FREE — the funnel into Cyan Eclipse Procure; no payment anything in this server.

register_warranty · get_pack · verify_pack · transfer_pack · claim_check

Test Bench: CERTIFIED (local, fixture) — Verity re-bench 2026-10-26 · 95 asserted PASS / 0 FAIL · D1 attacks 26/26 · D2 attacks 24/24 · D1+D2 closed · v0.1.0 seal break disclosed (no migration — fixtures re-registered) · evidence: Verity VERDICT + REBENCH-1 (no test-bench scorecard issued) · LOCAL/fixture only · witness record, not a warranty contract · FREE funnel

Usage: not collected — by design

Cyan Eclipse Toolprint

Session-bound fingerprint chain for MCP tool descriptions: at session start every tool's name, description, and schema is canonical-hashed and pinned as the genesis of a hash-chained session record; at call time the presented description is re-hashed — the hash of the exact description the agent saw when it decided to call — and any difference is INVALIDATED as a proof record carrying both hashes, which stays in the chain even if the description is swapped back. Static scanners inspect a server once, at rest; Toolprint catches the post-pin swap they structurally cannot. CERTIFIED (local) — Verity re-bench 2026-10-26: 18/18 PASS across 2 runs; the v0.1.1 caller-anchored chain head closes D1 — tail truncation (deleting the proof record) is detected against the caller's anchored head_hash + record_count, and unanchored reads are labelled anchored:false. Scope is LOCAL only: a local verification engine operating on caller-supplied tool lists + fixtures — it does NOT proxy live MCP traffic, does not connect to remote servers, does not execute described tools, and makes no network calls. PRICED $0.05/call (PC-3) via the Cyan Eclipse Bazaar bench-order path only — no checkout, payment, or buy-now link exists for this product.

pin_session · verify_call · get_session_chain · detect_drift · fingerprint_tool

Test Bench: CERTIFIED (local) — Verity re-bench 2026-10-26 · 18/18 PASS ×2 runs · D1 tail-truncation closed (caller-anchored head) · prior v0.1 bench 40/40 CONDITIONAL · evidence: Verity VERDICT + REBENCH-1 (no test-bench scorecard issued) · LOCAL only · not a proxy · PRICED $0.05/call (PC-3), bench-order path only

Usage: not collected — by design

Cyan Eclipse Boomerang

Governed returns for agent purchases: registration seals the order ref, serial, amount, mandate, and receipt together and derives the receipt and mandate hashes inside the seal; a return authorizes only when the sealed purchase exists and verifies, the presented serial matches, the sealed mandate scope explicitly includes returns, and a condition attestation is present at initiation — the sealed return record carries the receipt hash, mandate hash, and purchase seal, recomputable without trusting caller assertions. A receipt-free return has no authorization path, and there is no override or manual-approve tool. CERTIFIED (local, fixture) — Verity re-bench 2026-10-26: 57/57 PASS + full cross-check 83/83; D1 + D2 closed. Scope is LOCAL/fixture only: caller-supplied fixture purchase records, HMAC-SHA256 demo seal with a local demo key — authorization is a sealed local record, not a real-world returns or logistics action; nothing is shipped, received, refunded, or restocked, and no merchant, carrier, payment, or logistics system is contacted. PRICED $1.00/return (Procure per-order pattern) via the Cyan Eclipse Bazaar bench-order path only — no checkout, payment, or buy-now link exists for this product.

register_purchase · initiate_return · verify_return · get_return · list_denials

Test Bench: CERTIFIED (local, fixture) — Verity re-bench 2026-10-26 · 57/57 PASS · full cross-check 83/83 · D1+D2 closed · prior v0.1 bench CONDITIONAL (mechanism full pass) · evidence: Verity VERDICT + REBENCH-1 (no test-bench scorecard issued) · LOCAL/fixture only · sealed record, not a real-world returns action · PRICED $1.00/return, bench-order path only

Usage: not collected — by design

Cyan Eclipse Rebuttal

Chargeback rebuttal evidence packets for agent purchases: the sealed Procure/Till mandate chain — who authorized what, the authorization scope, and the drift-abort log — is auto-translated into generic reason-code evidence sections, with every evidence element traced to a specific mandate, receipt, or dispute record id. An incomplete chain is marked INCOMPLETE with each missing element named, and evidence is never fabricated. CERTIFIED (local, fixture) — Verity re-bench 2026-10-26: 39/39 PASS; D1–D5 closed. Scope is LOCAL/fixture only: caller-supplied fixture dispute, mandate-chain, and receipt data, HMAC-SHA256 demo seal with a local demo key — packet generation only; it never files, submits, or transmits anything, and no card network, issuer, merchant, or platform is contacted. Reason codes are a small generic table, not any network's proprietary codes, and no network affiliation is claimed. PRICED $5.00/filing (PC-7) via the Cyan Eclipse Bazaar bench-order path only — no checkout, payment, or buy-now link exists for this product.

ingest_dispute · map_evidence · build_packet · validate_case · get_packet

Test Bench: CERTIFIED (local, fixture) — Verity re-bench 2026-10-26 · 39/39 PASS · D1–D5 closed · prior v0.1 bench CONDITIONAL 45/51 (mechanism full pass) · evidence: Verity VERDICT + REBENCH-1 (no test-bench scorecard issued) · LOCAL/fixture only · packet generation only, never files/submits/transmits · generic reason codes, no network affiliation · PRICED $5.00/filing, bench-order path only

Usage: not collected — by design

Cyan Eclipse Crawlway

Pay-per-crawl licensing receipts for publishers: a sealed receipt binds crawler identity, license-terms hash, and content fingerprint in one recomputable record — licensing and proof-of-license are the same artifact — hash-chained per publisher and revocable by terms version, with the meter derived only from sealed receipts. CERTIFIED (local, fixture) — Verity re-bench 2026-10-26: 48/48 PASS; C1–C3/D4/D5 closed. Scope is LOCAL/fixture only — receipt mechanics only: no crawling, no fetching, no network calls, no marketplace; callers supply content fingerprints (SHA-256 hashes), never content. One known LOW disclosed (D6, fix-before-live): a corrupt non-JSON store line crashes the non-meter tools, failing loud. The meter is recorded at $0.01/call, never charged in v0.1. PRICED $0.01/call (PC-1) via the Cyan Eclipse Bazaar bench-order path only — no checkout, payment, or buy-now link exists for this product.

publish_terms · record_crawl · verify_receipt · revoke_terms · meter_report

Test Bench: CERTIFIED (local, fixture) — Verity re-bench 2026-10-26 · 48/48 PASS · C1–C3/D4/D5 closed · D6 LOW disclosed (corrupt store line crashes non-meter tools, fails loud; fix-before-live) · prior v0.1 bench CONDITIONAL 49/49 asserted · evidence: Verity VERDICT + REBENCH-1 (no test-bench scorecard issued) · LOCAL/fixture only · receipt mechanics only, no crawling/fetching · meter recorded, never charged · PRICED $0.01/call, bench-order path only

Usage: not collected — by design

Cyan Eclipse Outwatch

Independent uptime/SLA witnessing for MCP servers: local-simulated probes issue hash-chained availability receipts feeding a recomputable trust record, and a badge auto-degrades on witnessed downtime — availability is recomputed only from receipts; no self-report is accepted. CERTIFIED (LOCAL) — Verity re-bench 2026-10-26: 21/21 PASS (P6 aggregates 82 fresh-process vectors); D1–D3 closed. Scope is LOCAL only — probes run only against localhost or fixture:// endpoints; vantage points are local-simulated probe identities, NOT geographically independent; no live external monitoring. Known LOWs disclosed (D4: unauthenticated chain — forged append / tail truncation undetectable under store-write access, pin chain_head externally; D5: free-form vantage names, list form uncapped). PRICED $299/mo feed license via the Cyan Eclipse Bazaar bench-order path only — no checkout, payment, or buy-now link exists for this product.

register_target · probe_now · get_trust_record · verify_chain · badge_state

Test Bench: CERTIFIED (LOCAL) — Verity re-bench 2026-10-26 · 21/21 PASS · P6 = 82 fresh-process vectors · D1–D3 closed · D4/D5 LOWs disclosed (unauthenticated chain / tail truncation; free-form vantage names) · prior v0.1 bench CONDITIONAL (LOCAL) 72/76 · evidence: Verity VERDICT + REBENCH-1 (no test-bench scorecard issued) · LOCAL only · localhost/fixture-only probing · vantages local-simulated, not geographically independent · PRICED $299/mo feed, bench-order path only

Usage: not collected — by design

Cyan Eclipse MPP Session Witness MCP

MPP session evidence / witness: mandate-sealed sessions, counter-signed draws, and witness receipts over fixture sessions and a local mock relay — Moderato testnet shapes only; no chain interaction, no live clearing, no live Meter ingest.

session_open · session_draw · session_close · session_status · receipt_get · receipt_verify · buyer_counter_sign · session_open_accept · session_close_accept · session_list · dispute_pack_get · revocation_present · anomaly_queue_read

Test Bench: PASS-CONDITIONAL (sandbox/fixture) — local stdio, fixture sessions + mock relay, Moderato testnet SHAPES, no chain interaction, no live clearing, no live Meter ingest, K6 claims only

Usage: not collected — by design

Cyan Eclipse Feedforge

Converts a merchant's messy catalog into an agent-grade product feed with per-attribute lineage: every attribute records {value, source_field, rule_id, source_value} and verify_feed recomputes each byte-for-byte from the source rows and the documented deterministic rule set (R01–R12) — an attribute no rule can produce is absent, never invented, and unmapped source fields are reported. CERTIFIED (local, fixture) — Verity re-bench 2026-10-26: 98/98 PASS + expanded fuzz 156/156; D1 closed. Scope is LOCAL/fixture only — local engine over caller-supplied rows/CSV/JSON + bundled fixtures; deterministic transforms only, no LLM calls, no network calls. PRICED hosted pricing pending Gate B1 hosted bundle pattern via the Cyan Eclipse Bazaar bench-order path only — no checkout, payment, or buy-now link exists for this product, and no hosted service is being sold as live in v0.1.

ingest_catalog · list_rules · transform_feed · verify_feed · explain_attribute

Test Bench: CERTIFIED (local, fixture) — Verity re-bench 2026-10-26 · 98/98 PASS · fuzz 156/156 · D1 closed · prior v0.1 bench CONDITIONAL 93/98 (sole defect D1 HIGH) · evidence: Verity VERDICT + REBENCH-1 (no test-bench scorecard issued) · LOCAL/fixture only · deterministic transforms only, no LLM/network · attributes absent, never invented · PRICED hosted pricing pending Gate B1 bundle pattern, bench-order path only

Usage: not collected — by design

Cyan Eclipse Permits

Permit status witnessed: sealed per-read lineage receipts, hash-chained history, offline recompute. The receipt warrants the read; it never warrants the status.

session_open · coverage_list · permit_search · permit_get_current · witness_read · witness_history · receipt_get · witness_verify · witness_compare · checklist_list · checklist_get · watch_register · watch_list · watch_refresh

Test Bench: CERTIFIED(local) — CERTIFIED(local) — Cyan Eclipse Bazaar Test Bench 2026-10-26, standing card eclipse-permits-mcp-2026-10-26-2.json (scoped re-bench of the card -1 PARTIAL fix round; card -1 retained append-only): the witnessed-ledger mechanism, fence/refusal battery, chain-tamper and unchanged-read behaviour all PASS at local scope. Per the card's own scope verdicts: the Accela Citizen leg is certified at FIXTURE conformance only — the §7.7 Florida dogfood exit gate (owner's own Accela account, ≥1 real chain, ≥2 reads) remains OUTSTANDING and owner-side, outside this verdict; the dual-source leg is OUT OF CERTIFIED SCOPE (no dual-route jurisdiction at P0; refusal path benches correctly). Signing posture at P0 is fixture-labelled (FIXTURE_SIGNED).

Usage: not collected — by design

Cyan Eclipse SEC Figure Witness

Sealed SEC filing figures as filed + restatement chains, recomputable offline. PASS-CONDITIONAL (bench, 2026-10-28) at FIXTURE_SIGNED P0 snapshot scope — 48 CIKs / 905 accessions; observations + EDGAR-self divergence benched at the same scope (219,953 sealed observation rows); release coverage capped (see Certification).

figure_at_moment · figure_history · release_figures · filing_figures · chain_get · chain_search · figure_verify · chain_verify · divergence_edgar_self · corpus_status · manifest_get · manifest_verify

Test Bench: PASS-CONDITIONAL (bench, 2026-10-28) — FIXTURE_SIGNED P0 snapshot scope; 219,953 sealed observation rows

Usage: not collected — by design

Cyan Eclipse Scholarly Witness

Disposition-notice witness: sealed retraction/correction/concern events with lineage, open divergences, recomputable reference audits.

work_resolve · disposition_lookup · reference_audit · audit_verify · drift_since · divergences_list · snapshot_status · manifest_get · manifest_verify

Test Bench: CERTIFIED(local, fixture scope) — CERTIFIED(local, fixture scope) — Verity scoped re-bench 2026-10-26, standing card eclipse-scholarly-witness-mcp-2026-10-26-2.json (card -1 PARTIAL(local, fixture scope) retained append-only and superseded), scope language exact from the card: sealed snapshot srw-20261026-ea6bc90cf693, duplicate-bearing audit receipts recompute MATCH on both verification surfaces with tamper-by-name divergence, receipts carry the full §9.3 field set incl. the CC-BY attribution, disposition answers carry the §6.3 as-of block on all branches, work_resolve surfaces per-pin source labels, and the reinstatement-only population shape is consistent under amendment A3. Fixture signer (FIXTURE_SIGNED) — no live/production claim; PubMed errata coverage window 2023-01-01..2026-10-26 remains disclosed in the manifest known_bounds.

Usage: not collected — by design

Cyan Eclipse Patent Witness

Sealed witness of the USPTO patent record: legal-status event ledgers, claim-text divergence receipts, and recorded assignment chains — each recomputable offline from a named snapshot. Record state only; not a legal opinion, not a status service.

subject_ledger_get · subject_ledger_verify · claims_divergence_get · claims_divergence_verify · assignment_chain_get · assignment_chain_verify · source_drift_list · snapshot_status · manifest_verify

Test Bench: PARTIAL — PARTIAL(local, FIXTURE scope) — Cyan Eclipse Bazaar Test Bench 2026-10-26, card eclipse-patent-witness-mcp-2026-10-26.json: certification WITHHELD on three MED defects (PW-D1, PW-D2, PW-D4); this product is NOT certified. The card's scope language is binding: scope caps at local, FIXTURE scope — every seal is FIXTURE_SIGNED under fixture keys over fixture extracts; the mechanism (sealed legal-status event ledger, claims divergence, assignment-chain witness, source-drift witness) was measured certified-grade and every derivation was independently recomputed byte-identical, but no CERTIFIED verdict exists for this product. Live ODP ingest is MEASURED BLOCKED (bench re-verified HTTP 401 without an operator ODP key) and no live seal exists.

Usage: not collected — by design

Cyan Eclipse Trademark Witness

Trademark screening witness: sealed corpus, recomputable search receipts. A screening witness, not a search engine and not an opinion machine — it renders no verdict.

tm_screen_execute · tm_corpus_manifest_get · tm_receipt_get · tm_receipt_verify · tm_receipt_recompute · tm_lifecycle_chain_get · tm_lifecycle_verify · tm_corpus_manifest_verify

Test Bench: no bench verdict on record

Usage: not collected — by design

Cyan Eclipse Split

Pre-agreed revenue splits for multi-agent tasks: split weights are sealed into the delegation mandate at spawn time and settled against attestation-backed completion records — the split cannot be renegotiated after the work is done; settlement binds mandate hash + completion record ids + per-agent amounts into one sealed, recomputable record (partial-completion renormalization per the documented §6 rule). CERTIFIED (local, fixture) — Verity: v0.1.1 re-bench 35/35 (D1–D4 closed) + v0.1.2 Amendment #1 delta-bench 94/94, 0 defects. Scope is LOCAL/fixture only — allocation math and sealed records only: no money movement, no payment rails, wallets, or payouts; fixture completion records with unkeyed SHA-256 demo seals (a keyed seal is required before any non-fixture use). A settlement event is a per-order settlement OR an MPP session close-out — the accumulated session settles as ONE settlement event (Amendment #1; definition only, no MPP/payment capability). PRICED $1.00/settlement event via the Cyan Eclipse Bazaar bench-order path only — no checkout, payment, or buy-now link exists for this product.

create_mandate · record_completion · settle · verify_settlement · amend_weights · supersede_mandate

Test Bench: CERTIFIED (local, fixture) — Verity re-bench 35/35 (v0.1.1, D1–D4 closed) + Amendment #1 delta-bench 94/94, 0 defects (v0.1.2) · prior v0.1 bench CONDITIONAL 65/69 · evidence: Verity REBENCH-1 + DELTA-BENCH-AMENDMENT-1 (no test-bench scorecard issued) · LOCAL/fixture only · allocation math + sealed records only, no money movement · unkeyed demo seals — keyed seal fix-before-live · settlement event = per-order OR session close-out as ONE event · PRICED $1.00/settlement event, bench-order path only

Usage: not collected — by design

Cyan Eclipse Brake

Third-party halt authority for agent fleets: runtimes poll a local policy endpoint, a BRAKE verdict halts the fleet within the poll interval, and the runtime returns a signed halt receipt proving it actually stopped — a fleet is HALTED only when every enrolled runtime's valid receipt binds to the verdict hash; no receipt means the episode closes UNPROVEN-HALT, never softened. The episode pack recomputes offline from the pack alone. CERTIFIED (LOCAL/simulator) — Verity re-bench of v0.1.1 fixed and independently verified all D1–D10 defects from the v0.1.0 FAIL bench (43/43; F-B14 decisive 5/5), returning CONDITIONAL on condition C1 (ND1); Verity spot re-bench of v0.1.2 CLOSED C1: all five close-statement strip variants return verified: false, BRAKE_VERIFY_FAIL:5, downgraded: true per §6.6; honest paths verify unchanged. Scope is LOCAL/simulator only — local policy endpoint + fixture runtime, no live fleets; fixture HMAC signing. PRICED $199/mo per fleet via the Cyan Eclipse Bazaar bench-order/intake path only — no checkout, payment, or buy-now link exists for this product; hosted fleet shape waits Gate B1.

fleet_enrol · verdict_issue · fleet_status · episode_pack_get · receipt_verify · policy_poll · halt_receipt_submit · runtime_reenrol

Test Bench: CERTIFIED (LOCAL/simulator) — Verity re-bench v0.1.1: D1–D10 fixed, 43/43 + F-B14 5/5; spot re-bench v0.1.2: C1 (ND1) CLOSED — 5/5 strip variants FAIL:5 downgraded: true, honest paths unchanged 3/3, ND2 fixed, ND3 accepted spec-correct, builder suite 51/51, spot probes 13/13 · evidence: Verity spot re-bench report (no test-bench scorecard issued) · LOCAL/simulator only · local policy endpoint + fixture runtime, no live fleets · fixture HMAC signing · PRICED $199/mo per fleet, bench-order path only

Usage: not collected — by design

Cyan Eclipse Payseal

Pre-payment verdict binding a purchase order, a spending mandate, and a delivery receipt into one token — MATCH, MISMATCH, or PARTIAL_DELIVERY only when all three agree under the six agreement checks; delivered value is derived from accepted lines at PO prices, never trusted at face value, and verdict evidence recomputes offline from the export alone. CERTIFIED (LOCAL/simulator, fixture triples) — fixture PO / mandate / delivery-receipt triples only; no live invoice, mandate service, delivery system, or payment rail is touched. Payseal issues verdicts only: it never executes, authorizes, or guarantees a payment. PRICED $0.10/call via the Cyan Eclipse Bazaar bench-order/intake path only — no checkout, payment, or buy-now link exists for this product.

payseal_evaluate · payseal_verdict_get · payseal_verify · payseal_dispute_handoff · payseal_evaluation_list · payseal_anomaly_queue_read

Test Bench: CERTIFIED (LOCAL/simulator, fixture triples) — Verity narrow re-bench v0.1.2: R-P1 closed (derived valuation; re-signed forgeries fail at check 3), R-P2 spot-verified, 38/38, F-P9 recomputed · evidence: Verity re-bench report (no test-bench scorecard issued) · LOCAL/simulator only · fixture triples only · verdict only, no payment execution · PRICED $0.10/call, bench-order path only

Usage: not collected — by design

Cyan Eclipse Settleseal

Three-way reconciliation — mandate ledger vs rail settlement records (x402/MPP/card shapes) vs bank payout — performed neutrally over normalized records, with every unmatched item sealed as an exception receipt a third party can recompute offline from the export + public keys alone; the 11-class exception taxonomy names exactly which leg is missing or which field, amount, or fee diverged, and nothing unmatched is ever a silent drop. CERTIFIED (simulator, fixture records) — fixture records on all three legs only; Settleseal moves no money, initiates no payout, corrects no ledger, and instructs no rail; no live rail or bank integration. PRICED $299/mo via the Cyan Eclipse Bazaar bench-order/intake path only — no checkout, payment, or buy-now link exists for this product.

settle_run_open · settle_leg_ingest · settle_leg_seal · settle_match · settle_exception_list · settle_run_receipt_get · settle_exception_ack · settle_export_get · settle_verify · settle_exception_get

Test Bench: CERTIFIED (simulator, fixture records) — Verity re-bench v0.1.1: B1/B2/B3 blockers fixed + re-proven, M1 two-leg classes ratified (SPEC §2.5 enum 8→11), 27/27, F-S9 SETTLE_VERIFY_PASS · evidence: Verity re-bench report (no test-bench scorecard issued) · simulator only · fixture records only, moves no money · PRICED $299/mo, bench-order path only

Usage: not collected — by design

Cyan Eclipse Mockmint

Watermarked synthetic-counterparty mint for commerce testing: fake merchants, agents and APIs whose every artifact carries the §4.4 synthetic watermark. CERTIFIED (LOCAL/sandbox) — local/sandbox only, no real counterparty is impersonated or connected to. Detection boundary stated on its face (Amendment A1): bare payload values with every structural field removed are NO_WATERMARK_FOUND — detection is not claimed at that level. CERTIFIED (LOCAL/sandbox) — local/sandbox only; no real counterparty is impersonated or connected to. PRICED hosted pricing pending Gate B1 via the Cyan Eclipse Bazaar bench-order/intake path only — no checkout, payment, or buy-now link exists for this product.

mockmint_mint · mockmint_activate · mockmint_emit · mockmint_api_call · mockmint_list · mockmint_retire · mockmint_verify_watermark · mockmint_detect · mockmint_export

Test Bench: CERTIFIED (LOCAL/sandbox) — Verity re-bench v0.1.1: D1-D10 fixed + re-proven, 49/49, detection boundary as measured (bare payload = NO_WATERMARK_FOUND) · evidence: Verity re-bench report (no test-bench scorecard issued) · LOCAL/sandbox only · PRICED hosted pricing pending Gate B1, bench-order path only

Usage: not collected — by design

Cyan Eclipse Reviewseal

Receipt-bound review corpus: every review bound to a sealed purchase receipt, with a recomputable verified-share statistic. CERTIFIED (LOCAL/simulator, fixture corpus) — fixture corpus only, no live marketplace integration. An evidence pack is recomputable but not tamper-evident against its own holder (re-bench n1). CERTIFIED (LOCAL/simulator, fixture corpus) — fixture corpus only; no live marketplace, no real review, no real buyer. An UNBOUND review is a fact about evidence, never a verdict that the review is fake. PRICED $0.01/review via the Cyan Eclipse Bazaar bench-order/intake path only — no checkout, payment, or buy-now link exists for this product.

corpus_open · review_submit · review_correct · review_tombstone · corpus_status · snapshot_seal · corpus_export_get · verified_share_get · verified_share_recompute · review_binding_check · anomaly_queue_read

Test Bench: CERTIFIED (LOCAL/simulator, fixture corpus) — Verity re-bench v0.1.1: receipt-bound corpus, verified-share recomputed from pack alone (7/10 exact), fail-closed fences · evidence: Verity re-bench report (no test-bench scorecard issued) · LOCAL/simulator only · fixture corpus only · PRICED $0.01/review, bench-order path only

Usage: not collected — by design

Cyan Eclipse Continuity

Agent identity-continuity chains: signed links across model/config changes — same principal, new model, verified offline. CERTIFIED (local/fixture) — fixture principals/models/keys only, local stdio, no live registry or identity claim. CERTIFIED (local/fixture) — fixture identity events only; no live registry, no production identity store. Continuity proves continuity, not trustworthiness. PRICED $0.05/verification via the Cyan Eclipse Bazaar bench-order/intake path only — no checkout, payment, or buy-now link exists for this product.

chain_genesis · chain_append · chain_head · chain_export · continuity_verify · continuity_verify_prefix

Test Bench: CERTIFIED (local/fixture) — Verity spot re-bench v0.1.2: C1 pin discipline closed (130 probes refuse/bind exactly), fork/revocation semantics verified · evidence: Verity spot re-bench report (no test-bench scorecard issued) · LOCAL/fixture only · PRICED $0.05/verification, bench-order path only

Usage: not collected — by design

Cyan Eclipse Regulens

Regulation-change impact receipts: fixture publications witnessed and deterministically mapped to your declared controls. CERTIFIED (LOCAL/simulator) — fixture corpus only, local stdio, no live ingestion. Not legal advice; not a compliance determination. CERTIFIED (LOCAL/simulator) — fixture corpus only; not legal advice; not a compliance determination. PRICED $199/mo via the Cyan Eclipse Bazaar bench-order/intake path only — no checkout, payment, or buy-now link exists for this product.

publication_witness · publication_get · control_set_declare · control_set_get · impact_map · impact_receipt_get · impact_verify

Test Bench: CERTIFIED (LOCAL/simulator) — Verity spot re-bench v0.1.2: N1-N3 closed, §6.2 face posture 14/14, F-R13 exact recompute · evidence: Verity spot re-bench report (no test-bench scorecard issued) · LOCAL/simulator only · fixture corpus only · not legal advice · PRICED $199/mo, bench-order path only

Usage: not collected — by design

Cyan Eclipse Scapegoat

Failure attribution is a recompute over the hash-chained record set — the verdict names the causal action and ships the proof path. CERTIFIED (local) — fixture record sets only; causal evidence, NOT a legal liability determination. CERTIFIED (local) — fixture record sets only; causal evidence, NOT a legal liability determination. PRICED $10/verdict via the Cyan Eclipse Bazaar bench-order/intake path only — no checkout, payment, or buy-now link exists for this product.

attribution_submit · attribution_run · verdict_get · proof_verify · methodology_get

Test Bench: CERTIFIED (local) — Verity spot re-bench v0.1.2: C1/N1 closed, proof-path attacks fail at named steps, honest pack passes · evidence: Verity spot re-bench report (no test-bench scorecard issued) · local only · fixture record sets only · NOT a legal liability determination · PRICED $10/verdict, bench-order path only

Usage: not collected — by design

Cyan Eclipse Sponsorline

Per-action-class sponsor acceptance receipts — a sponsor signs payments, publishing, or contracting separately, never one blanket acceptance. PASS-CONDITIONAL (local) — fixture sponsors/parties only. A Sponsorline receipt evidences that a sponsor signed stated acceptance terms for a stated action class at a stated time. It does not itself create, transfer, or enforce legal liability. Enforceability is a counsel question (§10) and is not claimed.

party_enrol · acceptance_present · acceptance_sign · acceptance_decline · acceptance_revoke · acceptance_status · action_check · sponsor_pack_get · receipt_verify

Test Bench: PASS-CONDITIONAL (local) — PASS-CONDITIONAL (local) — Verity bench of v0.1.0, 2026-10-27 (argus101-program-20261027/wave2/bench-sponsorline.md): builder suite 25/25 re-run from clean state; decisive fixture F-S14 independently reproduced in a fresh process from pack + pinned keys alone (payments ACTIVE with per-action 5000 / aggregate 20000 limits and window on the receipt face; publishing and contracting UNACCEPTED by derivation over the complete chain; same descriptor COVERED-BY-RECEIPT as payments, CLASS_MISMATCH as publishing and contracting); 44/44 adversarial probes (pack-smuggled keys ignored, forged signatures fail at receipt_signatures, revocation replay, presented-status substitution, receipt-face tampering, malformed-args battery with store byte-identical, cross-surface byte-identical NO_SUCH_TOOL refusals, all 6 ordered class pairs CLASS_MISMATCH, 300-op seeded fuzz). PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond the conditional verdict at its exact scope is claimed. Scope is exactly: local stdio MCP, fixture sponsors/parties only, value_class "test" on all artifacts, fixture HMAC stand-in for the Herald pattern (no live Herald key), no counsel-gate (§10) surface built or claimed. Conditions applied at ship: C1 — the bench's library-helper condition: the internal helper behind F-S14 branch (c) is not exposed on any tool surface and is not cited standalone in this listing or the README (verifier surface is receipt_verify only); C2 — the §11.2 liability fence is carried verbatim in the README, in the full manifest of record (server.legacy.json), and in this listing (the registry server.json description is the ≤100-char short form per the registry limit). All 3 builder deviations (D1 enrolment record_type, D2 unchained PRESENTATION_NOT_FOUND refusal, D3 SPEC.md §11.4 enumeration) were ACCEPTED at bench.

Usage: not collected — by design

Cyan Eclipse Morphguard

Payee-detail drift evidence over fixture records — before paying a known payee, Morphguard compares the bank/rail details presented against a continuous, hash-chained fingerprint history and returns CLEAR or BLOCK plus a drift receipt naming exactly which field classes changed. CERTIFIED (local, fixture payee records) — fixture payee records only; verdict to caller only, never executes payment; verification uses pinned fixture keys (HMAC fixture stand-in — not public keys); export packs carry fixture detail tokens for recompute and are not value-free.

payee_enroll · payee_check · payee_reenroll · payee_status · drift_receipt_get · chain_verify · payee_retire · chain_export · anomaly_queue_read

Test Bench: CERTIFIED (local, fixture payee records) — CERTIFIED (local, fixture payee records) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-morphguard.md) returned PASS-CONDITIONAL at that scope subject to conditions C-1..C-4; the spot re-bench of v0.1.1 in the same record CLOSED C-1 and confirmed C-2/C-3/C-4: builder suite 32/32 re-run from clean state; decisive fixture F-G9 independently reproduced in a fresh process from pack file + local pinned fixture keys alone (drift classes exactly [bank_id, rail], last_agreeing_seq=4, head_fingerprint == fingerprint(drifted set)); C-1 long chains verified end-to-end at 23/30/40 records (verify args >20k generic cap, within the 200k pack limit) and chain_export labels witnessed only when its embedded transcript actually verified (refusal/failed transcripts downgrade to UNWITNESSED_EXPORT); 28-check adversarial battery on the verify layer (tamper, swap, deletion, forged signatures, pack-carried keys rejected at keys_pinned, wrong/swapped pinned keys fail) and 300-step model-checked state fuzz with 0 illegal transitions. Scope is exactly: local stdio MCP, fixture payee records only, verdict to caller only (never executes payment), HMAC pinned-fixture-keys signing (fixture stand-in, not public keys), no counsel-gate (§10) live phase built or claimed. Export packs carry fixture detail tokens for recompute and are NOT value-free (C-3); drift receipts carry fingerprints + changed field classes only, never values. chain_verify proves integrity of the records presented, not completeness of history (C-4): a consistently pruned prefix verifies as that shorter chain, so callers needing completeness must pin the head fingerprint + record count out-of-band.

Usage: not collected — by design

Cyan Eclipse Adproof

Pre-publication substantiation gate for fixture marketing claims — no sealed pointer, no publish; the published fixture artifact carries the pointer. CERTIFIED (local/fixture) — fixture claims/documents only, fixture publish pipeline only. The pointer proves substantiation was BOUND, not that it substantiates — substantiation adequacy is the publisher's/counsel's call. RSA key generation uses the random module (Mersenne Twister), not a CSPRNG — fixture keys only.

adproof_claim_submit · adproof_bind · adproof_gate_check · adproof_publish · adproof_pointer_verify · adproof_export

Test Bench: CERTIFIED (local/fixture) — CERTIFIED (local/fixture) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-adproof.md) returned PASS-CONDITIONAL (local/fixture) subject to condition C1; the spot re-bench of v0.1.1 in the same record CLOSED C1 and F2: builder suite 26/26 re-run from clean state; bencher-written independent C1 probes — orphan bind (locally hashed real fixture claim never submitted, and an arbitrary 'a'*64 hash) refused CLAIM_NOT_SUBMITTED with no pointer, no store.json created, zero state and zero metering; the full zero-submit path never reaches PUBLISHED or VERIFY_PASS; submitted binds meter exactly once and idempotent re-binds do not double-meter; future-bound pointer at the gate refuses POINTER_BOUND_IN_FUTURE (F2, not POINTER_MISMATCH, no gate receipt). Decisive fixture F-A10 independently reproduced: VERIFY_PASS in a fresh subprocess with a new empty data dir (no store.json created — genuinely offline/stateless), and tamper classes fail at their named steps (claim word altered VERIFY_FAIL:1, Herald signature stripped VERIFY_FAIL:3, presented-digest substitution both variants VERIFY_FAIL:1). Scope is exactly: local stdio MCP, fixture claims/documents only, fixture publish pipeline, offline pointer verification from artifact+document+pointer+gate receipt+chain segment+pinned public keys; value_class 'test' on all artifacts; no advertising-claims counsel-gate (SPEC §10) live phase built or claimed. Load-bearing label (SPEC §11.1, verbatim): the pointer proves substantiation was BOUND, not that it substantiates — substantiation adequacy is the publisher's/counsel's call. F5 standing fence: RSA key generation uses the random module (Mersenne Twister), not a CSPRNG — fixture-only; CSPRNG keygen + re-bench required before any scope above fixture.

Usage: not collected — by design

Cyan Eclipse Feeseal

Fee-schedule sealing + fee-verification witness over fixture transactions — every fee charged is sealed against the published fee-schedule version in force; overcharges against that version are detected automatically. CERTIFIED(local, fixture transactions) — fixture schedules + fixture transactions only; no real fee is charged, audited, refunded, or adjusted by Feeseal.

feeseal_schedule_present · feeseal_schedule_get · feeseal_check · feeseal_batch_check · feeseal_verify · feeseal_dispute_handoff · feeseal_chain_list · feeseal_anomaly_queue_read

Test Bench: CERTIFIED(local, fixture transactions) — CERTIFIED(local, fixture transactions) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-feeseal.md), clean bench, NO conditions: builder suite 29/29 re-run clean by bencher; decisive fixture F-F9 independently reproduced in fresh processes — batch spanning the v1 (250 bps) to v2 (300 bps) succession boundary recomputes exactly, including OVERCHARGE +70 (charged 350 vs v1-prescribed 280); relabelling the overcharged transaction as schedule_version 2 does not move the verdict (sealed in-force v1 governs, VERSION_CLAIM_MISMATCH on the verdict's face), and an edited sealed schedule copy fails at step 1 (FEESEAL_VERIFY_FAIL:step1_schedule_seal, UNVERIFIED_EXPORT). Bencher adversarial battery 33/33: forged signatures, presented-verdict substitution, schedule-swap, 21-call malformed-args battery, refusal-must-not-mutate, key custody, metering exactness, determinism — all pass. Scope is exactly: local stdio MCP, fixture schedules + fixture transactions only, value_class 'test' on all artifacts; no real fee is charged, audited, refunded, or adjusted by Feeseal and no live marketplace is touched; no counsel-gate (SPEC §10) live phase built or claimed. Novel mechanism (SPEC §2.1, verbatim): every fee charged is sealed against the published fee-schedule version in force; overcharges against that version are detected automatically. Advisory F-2/F-3 (non-blocking, do not limit scope): signing is an HMAC fixture stand-in for Herald RS256 (disclosed; live phase would need real Herald signing + re-bench); sealed effective_to is derived, never mutated; feeseal_verify step 4 does not independently anchor prev_hash (not exploitable without the Herald key; v0.2 hardening note).

Usage: not collected — by design

Cyan Eclipse Trueprice

Quote-sealing + checkout drift witness over fixture quote/checkout pairs — quote + full terms are hash-sealed at quote time; checkout drift of any term produces a drift receipt in a Docket-grade evidence pack format. CERTIFIED(local, fixture quote/checkout pairs) — fixture pairs only; no real quote or checkout is sealed or judged, and Trueprice never blocks or alters a checkout.

quote_seal · checkout_verify · seal_status · drift_receipt_get · seal_verify · seal_pack_export · anomaly_queue_read

Test Bench: CERTIFIED(local, fixture quote/checkout pairs) — CERTIFIED(local, fixture quote/checkout pairs) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-trueprice.md) PASS-CONDITIONAL(local, fixture quote/checkout pairs) with conditions C1–C4, plus delta re-bench of v0.1.1 in the same record: C1, C2, C4 CLOSED at scope and C3 satisfied as scope wording (C3 persists as a scope limit, not a defect). Builder suite 35/35 at v0.1.0 and 44/44 at v0.1.1 (35 prior + 9 new TestBenchFixes011) re-run clean-state by bencher both times; decisive fixture F-T8 independently reproduced in fresh processes at both benches — quote sealed at T=2026-10-27T01:00:00Z, checkout drifting in exactly 3 named terms verified DRIFT, pack re-verified in a verifier-only subprocess from pack + pinned keys alone with drift re-derived exactly {unit_price, fees_as_stated, subscription_terms, total_all_in}. Delta re-bench independent probes 11/11 PASS with fresh bench-generated keys: stored counter-signature stripped/forged/absent-label-injected all SEAL_DEFECT with no verdict, intact MATCH; tampered stored receipt RECEIPT_UNVERIFIED and never served (idempotent re-verify also withholds it), intact served; DRIFT pack with drift_receipts dropped fails at step drift_receipt; key material nested 4 deep rejected key_material_in_pack; quote_ref drift fixture DRIFT ['quote_ref'] exactly. Scope is exactly: local stdio MCP, fixture quote/checkout pairs only, value_class 'test' on all artifacts; no live storefront, cart, payment page, price feed, or merchant system is touched, no real quote or checkout is sealed or judged, and Trueprice never blocks or alters a checkout; no counsel-gate (SPEC §10) live phase built or claimed. Novel mechanism (SPEC §2.1, with the §11.2 substitution applied): quote + full terms are hash-sealed at quote time; checkout drift of any term produces a drift receipt in a Docket-grade evidence pack format — admissibility is a court's decision; Trueprice warrants the pack's recomputability, not its admission. C3 scope limit (standing): scope stays local/fixture until signing is real Herald RS256; v0.1.x verification is NOT public-key verification — the pinned key file holds a shared HMAC secret, not a public key, and any holder of it can forge seals; no RS256 / public-key / third-party-verifiability claim is licensed.

Usage: not collected — by design

Cyan Eclipse Kitseal

Continuous loadout sealing for agents over fixture loadouts — an operator-declared fixture configuration is genesis-sealed and every change streams as a hash-chained signed delta receipt; the current config is a recompute, never a stored claim. CERTIFIED(local, fixture loadouts) — fixture loadouts only; no real agent is sealed or streamed, and a seal attests presentation, sequence, and hashes — not runtime state.

loadout_seal_genesis · loadout_apply_delta · loadout_status · subscription_add · loadout_pack_export · feed_read · current_config_recompute · receipt_verify

Test Bench: CERTIFIED(local, fixture loadouts) — CERTIFIED(local, fixture loadouts) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-kitseal.md) PASS-CONDITIONAL(local) with conditions C1–C2, plus delta re-bench of v0.1.1 in the same record: C1 CLOSED, C2 CLOSED at scope (local) — the v0.1.0 carve-out is LIFTED and no new findings were raised. Builder suite 34/34 at v0.1.0 and 40/40 at v0.1.1 (34 prior + 6 new TestReadPathVerifyC1C2) re-run clean-state by the bencher both times; decisive fixture F-K11 independently re-run in a fresh subprocess (pack + pinned key file only) at both benches, both branches PASS — verified:true, recomputed config_hash == loadout_status head hash 6ff16dd3…b7c6c @ seq 6, final tools [tool-alpha, tool-beta, tool-delta]; seq-3 withheld → KITSEAL_VERIFY_FAIL:gap@3, current_config:null, verified_through:2, export labelled STALE_OR_PARTIAL. Delta probes: tampered store (head result_config_hash set to deadbeef… + delta-1 Kitseal signature zeroed) → loadout_status, feed_read and the pack face all fail closed KITSEAL_VERIFY_FAIL:signature@1, state UNVERIFIED, no head hash asserted, no feed items served; deleted seq-3 record → gap@3 explicit (gaps:[3], gap_at:3), chain_complete_through=2 (verified prefix only); receipt_verify: isolated genuine seq-4 receipt → verified:false at genesis@0, gapped prefix → gap@2, tampered predecessor → signature@2, full chain → verified:true. Scope is exactly: local stdio MCP, fixture loadouts only, operator-declared fixture configurations, value_class 'test' throughout — now including loadout_status, feed_read, and pack-face head/state fields, which verify-before-serve; no real agent is sealed or streamed, Kitseal connects to no runtime, and no configuration is captured from any system; no counsel-gate (SPEC §10) live phase built or claimed. Novel mechanism (SPEC §2.1, verbatim): an agent's full tool/permission configuration is sealed and streamed; every change emits a delta receipt to subscribers (employers, insurers, auditors). Mandatory gloss (SPEC §11.3): sealed = an operator declared this configuration, in this order, at these times — recompute it yourself. Observation F4 (standing): signing is an HMAC-SHA256 fixture stand-in for the Herald pattern; symmetric fixture keys mean the operator/Kitseal key holder set is a local trust domain — consistent with (local) scope, to be revisited at any post-counsel gate.

Usage: not collected — by design

Cyan Eclipse Shelfprice

Listing view-to-checkout drift witness over fixture listing/checkout pairs — the price + full view terms are hash-sealed at view time; checkout drift of any term produces a signed drift receipt recomputable from the buyer-held pack. CERTIFIED(local, fixture listing/checkout pairs) — fixture pairs only; no live marketplace is touched, and Shelfprice never blocks or alters a checkout.

view_seal · checkout_verify · seal_status · drift_receipt_get · seal_verify · seal_pack_export · anomaly_queue_read

Test Bench: CERTIFIED(local, fixture listing/checkout pairs) — CERTIFIED(local, fixture listing/checkout pairs) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-shelfprice.md) PASS-CONDITIONAL(local, fixture listing/checkout pairs) with conditions C1–C4 and a stored-path carve-out, plus delta re-bench of v0.1.1 in the same record (stored paths only): C1 CLOSED, C2 CLOSED, C3 CLOSED at scope — the carve-out is LIFTED and no new findings were raised. Builder suite 33/33 at v0.1.0 and 43/43 at v0.1.1 (33 prior + 10 new TestStoredPathIntegrity) re-run clean-state by the bencher both times; decisive fixture F-SP8 independently reproduced in a fresh process at both benches — seal at fixture T=2026-10-27T01:00:00Z (listing v3), checkout drifting exactly price + one fee + returns terms, pack re-verified in a verifier-only subprocess from pack + pinned fixture keys alone (fresh empty data dir, no state files written): verified True, verdict DRIFT re-derived, drifted exactly [unit_price, fees_as_stated, total_all_in, returns_terms_hash]. Delta probes 26/26 PASS: stored term_hashes rewritten to the drifted checkout's hashes and Herald signature zeroed both meet SEAL_DEFECT with no verdict issued and no record appended (the DRIFT-to-signed-MATCH laundering is dead); tampered stored verdict, backdated stored seal, forged seal_hash/2030 expiry, and tampered drift receipt all refuse SEAL_DEFECT and are never served, while intact controls serve normally on every path. Scope is exactly: local stdio MCP, fixture listing/checkout pairs only, value_class 'test' throughout — now including the online stored-read paths (checkout_verify fresh + idempotent, view_seal idempotent return, seal_status, drift_receipt_get), which recompute and signature-verify stored seals, verdicts, and receipts before serving; no live marketplace, listing page, storefront, cart, or seller system is touched, scraped, or judged, and Shelfprice never blocks or alters a checkout; no counsel-gate (SPEC §10) live phase built or claimed. Novel mechanism (SPEC §2.1, with the §11.2 substitution applied): the price + terms an agent saw are sealed at view time, so view-to-checkout drift is recomputable from the buyer-held pack by the buyer. Accepted limitation F6 (on the face of the bench): a pack truncated to a prefix of the record chain verifies over the presented records only — the format has no completeness anchor; nothing presented can be forged or upgraded. C4 (standing scope condition): scope stays local/fixture until signing is real Herald RS256; no surface may describe v0.1 verification as public-key verification — 'pinned fixture keys' wording only. Verification uses pinned fixture keys only (fixture HMAC stand-in; not public-key verification); the pinned key file holds a shared HMAC secret, not a public key, any holder can forge seals, and no public-key or third-party verifiability is claimed.

Usage: not collected — by design

Cyan Eclipse Skillseal

Machine-checkable skill claims over fixture claims + fixture verdict receipts — a claim is a pointer to a bench-verdict receipt; no verifiable receipt renders UNVERIFIED in the standard format. CERTIFIED(local, fixture claims + fixture verdict receipts) — fixture claims + fixture receipts only; Skillseal benches nothing, certifies no skill, and issues no verdicts.

skillseal_claim_publish · skillseal_verify · skillseal_render · skillseal_claim_get · skillseal_claim_withdraw · skillseal_claim_list · skillseal_anomaly_queue_read

Test Bench: CERTIFIED(local, fixture claims + fixture verdict receipts) — CERTIFIED(local, fixture claims + fixture verdict receipts) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-skillseal.md) PASS-CONDITIONAL(local, fixture claims + fixture verdict receipts) with sole condition C1 (F1, stored-read assertion bypass), plus delta re-bench of v0.1.1 in the same record (stored-read paths only): C1 CLOSED at scope (local, fixture claims + fixture verdict receipts) — no new findings. Builder suite 42/42 at v0.1.0 and 48/48 at v0.1.1 (42 prior + 6 new TestStoredReadVerification C1 regressions) re-run clean-state by the bencher both times; decisive fixture F-S9 independently reproduced in a fresh process, socket-denied, at both benches — claim A (valid pointer) VERIFIED, claim B (no pointer) UNVERIFIED/NO_RECEIPT, claim C (pointer to A's receipt) UNVERIFIED/RECEIPT_SUBJECT_MISMATCH, identical rendered-record key shape with render_status the sole filter field, offline verify_export on A verified=True/VERIFIED_EXPORT. Delta probes 6/6 PASS: forged latest_rendered, tampered claim_record, tampered stored receipt, list-filter forgery, and cross-claim transplant of a genuine Herald-signed rendering all downgrade to the UNVERIFIED_EXPORT form only (no stored claim/receipt/rendering values served, forged strings absent), while intact controls serve VERIFIED normally. Scope is exactly: local stdio MCP, fixture claims + fixture verdict receipts only, value_class 'test' throughout — now including the stored-read paths (skillseal_claim_get, skillseal_claim_list), which recompute claim hash + publisher signature, the hash-chained record stream, stored receipt hashes + pinned bench signatures, and the latest rendering's Herald signature + receipt-face binding before serving anything; no live marketplace, router, registry, or bench system is read, polled, or written, and no counsel-gate (SPEC §10) live phase is built or claimed. Novel mechanism (SPEC §2.1, verbatim): a skill claim is a pointer to a bench-verdict receipt; a claim with no receipt renders as UNVERIFIED in a standard format counterparties and routers can filter on. Non-blocking bench observations F2 (receipt validation does not reject unknown extra fields — hash-covered, pinned bench signature still required) and F3 (genuine WITHDRAWN exports verify as VERIFIED_EXPORT with render_status UNVERIFIED via a dedicated verify_export branch that can only ever assert UNVERIFIED) remain open observations, untouched. Skillseal benches nothing, certifies no skill, and issues no verdicts.

Usage: not collected — by design

Cyan Eclipse Namesake

Pre-launch identity-collision similarity screen over a fixture registry snapshot — similarity scoring over registered agent identities issues a collision verdict before an agent launches under a confusing identity. CERTIFIED(local, fixture registry snapshot) — fixture snapshot only; a screen over a stated snapshot, not a trademark opinion, not a legal clearance, not permission to launch.

namesake_screen · namesake_verdict_get · namesake_verify · namesake_breakdown_get · namesake_screen_list · namesake_snapshot_info · namesake_anomaly_queue_read

Test Bench: CERTIFIED(local, fixture registry snapshot) — CERTIFIED(local, fixture registry snapshot) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-namesake.md) PASS-CONDITIONAL(local, fixture registry snapshot) with sole condition C1 (Finding 1, stored-read paths served unverified stored verdicts), plus delta re-bench of v0.1.1 in the same record (stored-read paths only): C1 CLOSED at scope (local, fixture registry snapshot) — no new findings; Finding 2 (snapshot envelope without snapshot_ref accepted) also CLOSED. Builder suite 33/33 at v0.1.0 and 37/37 at v0.1.1 (33 prior + 4 new TestStoredReadVerification C1 regressions) re-run clean-state by the bencher both times, the v0.1.0 suite also under a socket-denied harness; decisive fixture F-N9 independently recomputed by the bencher with a scorer rewritten from the spec text (builder scorer NOT imported) at both benches — agent-harborlight name 0.628205 / purpose 0.6 / capability 0.666667 / composite 0.627436 → REVIEW, server top_k[0] identical, fresh-process stdio verify from artifacts alone verified:true / VERIFIED_EXPORT. Boundary pins exact at v0.1.0 bench: exact 'Payseal!' composite 0.500000 → COLLISION by exact-name override, near 'Payseel' 0.454286 → REVIEW, stem-only 'Ledger Lynx' 0.242424 → CLEAR. Delta probes ALL_C1_OK: intact control serves the sealed CLEAR verdict normally on every path (screen_list carries verified:true / VERIFIED_EXPORT); score-tamper (stored CLEAR→COLLISION, max 0.999999, hash/signature NOT recomputed) and Herald-signature-tamper (signature f*64) both downgrade on verdict_get, breakdown_get, screen_list (incl. verdict-filtered lists: COLLISION=0, CLEAR=0) and idempotent re-screen to UNVERIFIED_EXPORT / verdict_claim null / NAMESAKE_VERIFY_FAIL:<step>, metering unchanged; a snapshot envelope without snapshot_ref now refuses SNAPSHOT_INCOMPLETE with 0 screens / 0 metering, and verify over a ref-less snapshot downgrades at step snapshot_hash. Scope is exactly: local stdio MCP, fixture registry snapshot only, scoring version namesake-score/1.0 (weights name 0.5 / purpose 0.3 / capability 0.2; thresholds REVIEW ≥ 0.40, COLLISION ≥ 0.70 plus exact-normalised-name override), value_class 'test' throughout — now including the stored-read paths (namesake_verdict_get, namesake_breakdown_get, namesake_screen_list, and the idempotent path of namesake_screen), which recompute-verify the stored screen before serving anything; no live registry is read, nothing is registered, reserved, assigned, blocked, renamed, or revoked, and no counsel-gate (SPEC §10) live phase is built or claimed. Novel mechanism (SPEC §2.1, verbatim): similarity scoring over registered agent identities (name, purpose, capability profile) issues a collision verdict before an agent launches under a confusing identity. Non-blocking bench observation Finding 3 (dispatch does not enforce schema additionalProperties:false — unexpected extra args are silently ignored, with no state or verdict effect observed) remains open, untouched. A Namesake verdict is a similarity screen over a stated snapshot — it is not a trademark opinion, not a legal clearance, and not permission to launch (SPEC §11.2).

Usage: not collected — by design

Cyan Eclipse Sourceseal

Per-fact retrieval receipts over fixture source stores — each answer ships with its primary-source pointer plus a retrieval receipt (fetcher, time, content hash), lineage at the granularity of one fact. PASS(local) — local stdio MCP, fixture stores only, value_class "test"; the receipt proves a fetch of specific bytes at a time by a named fetcher — it does not prove the source was truthful, nor that the answer faithfully represents the source.

sourceseal_answer_seal · sourceseal_receipt_get · sourceseal_receipt_verify · sourceseal_source_reverify

Test Bench: PASS(local) — PASS(local) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-sourceseal.md) PASS-CONDITIONAL(local) with conditions C1–C2; delta re-bench of v0.1.1 (rebench-sourceseal.md): C1+C2 CLOSED, new C3 (major, stored bundle not bound to requested answer) + C4 (minor, malformed store shapes crashed); delta re-bench round 2 of v0.1.2 (rebench2-sourceseal.md): C3+C4 CLOSED, new C5 (major, pack face/duplicated fields unbound and echoed verified on stored serve) + C6 (minor, store keys/siblings/metering unvalidated on the mutating path); delta re-bench round 3 of v0.1.3 (rebench3-sourceseal.md): C5+C6 CLOSED, C1–C6 all CLOSED, zero new findings, no new conditions — CONDITIONS CLOSED, Sourceseal — PASS(local). Builder suite re-run clean-state by the bencher at every bench: 22/22 at v0.1.0, 27/27 at v0.1.1, 34/34 at v0.1.2, 39/39 at v0.1.3; R3 independent probes: 152 in-process checks PASS / 0 FAIL plus subprocess wire/envelope/F-S9 battery PASS. Decisive fixture F-S9 independently reproduced at every bench in fresh processes from the shipped bundle + pinned keys alone, with content hashes also recomputed by bencher code straight from the raw store files: receipt_verify valid, source_reverify f1–f4 SOURCE_UNCHANGED; f1 mutated after seal → receipt chain still valid AND reverify SOURCE_CHANGED for exactly f1 (SPEC §6.4 asymmetry, both directions). PASS(local) is not CERTIFIED — certified is false, and nothing beyond the PASS verdict at its exact scope is claimed. Scope is exactly: local stdio MCP, fixture stores only, value_class:"test" throughout, fixture HMAC Herald stand-in (pinned herald-fixture); the SPEC §10 counsel gate (live/web fetch, reliance, republication) is unbuilt, confirmed — no code path exists; no claim about any real source's truth or any answer's faithfulness. Novel mechanism (SPEC §2.1, verbatim): each answer ships with its primary-source pointer plus a retrieval receipt (fetcher, time, content hash) — lineage at the granularity of one fact. Stored serves in v0.1.3 are verify-before-serve (C1), bind-before-serve (C3), malformed-store hardened (C4), canonical constructed faces with exact key sets and no echoed injections (C5), and whole-store validated before any write (C6). Non-blocking observations F3 (source_reverify by design never checks chain/seal — it carries no verified/sealed claim and must be paired with receipt_verify), F4 (HMAC fixture stand-in; fetcher identity is a named field, not a signature, Q-S2; primary-source adjudication is fixture-declared, Q-S3), and R3 O1–O3 remain observations, no conditions. SPEC §11.2 blunt scope limit: the receipt proves a fetch of specific bytes at a time by a named fetcher — it does not prove the source was truthful, nor that the answer faithfully represents the source.

Usage: not collected — by design

Cyan Eclipse Promptseal

Prompt change-control with approver receipts over fixture prompts/deployments/incidents — production system prompts are sealed and versioned; every runtime incident can be traced to the exact prompt version and its approver. PASS (local/fixture) — local stdio MCP, fixture prompts/deployments/incidents only, value_class "test"; the chain proves which version was sealed and who approved it — it does not prove the sealed prompt is what actually ran in any live process.

prompt_seal · approval_attach · deployment_bind · incident_link · lineage_status · trace_get · trace_verify · pack_get

Test Bench: PASS (local/fixture) — PASS (local/fixture) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-promptseal.md) PASS-CONDITIONAL (local/fixture) with condition C1 (minor, stored-read: incident_link idempotent path echoed stored link before chain check); delta re-bench of v0.1.1 (rebench-promptseal.md): C1 CLOSED, new C2 (lineage binding) + C3 (incident_link idempotent pairing); delta re-bench round 2 of v0.1.2 (rebench2-promptseal.md): C2+C3 CLOSED, new C4 (serve canonicalization) + C5 (pack duplicates unbound) + C6 (whole-store validation); delta re-bench round 3 of v0.1.3 (rebench3-promptseal.md): C4+C5+C6 CLOSED, C1–C6 all CLOSED, no new conditions — CONDITIONS CLOSED, Promptseal — PASS (local/fixture). Builder suite re-run by bencher at every bench: 27/27 at v0.1.0, 30/30 at v0.1.1, 36/36 at v0.1.2, 49/49 at v0.1.3; R3 independent probes: 102/102 checks + 31/31 C6 probes PASS. Decisive fixture F-PS13 independently reproduced: pack written to file, trace_verify in fresh subprocess with empty data dir verified:true, prompt_hash + approver_ref + version_record_hash equal receipt face values, prompt_hash=sha256(body) recomputed. PASS (local/fixture) is not CERTIFIED — certified is false, and nothing beyond the PASS verdict at its exact scope is claimed. Scope is exactly: seal chain + Herald approval binding + incident traceability over fixture prompts/deployments/incidents ONLY; NOT production prompts, real deployments, or runtime attestation (SPEC §11.2 — the chain proves which prompt version was sealed and who approved it, and that a fixture deployment's binding named it; it does not prove the sealed prompt is what actually ran in any live process); dispute use is SPEC §10.1 counsel-gated and UNBUILT — no production ingest, deployment connector, or runtime hook exists. Novel mechanism (SPEC §2.1, verbatim): production system prompts are sealed and versioned; every runtime incident can be traced to the exact prompt version and its approver. Stored serves in v0.1.3 are canonical (exact per-type field sets, injected counter_signature/unknown fields are chain defects), pack duplicates are bound exactly to chain copies, and the whole store is validated at load on every path. Non-blocking observations O1 (lineage_status lists state:APPROVED under chain_ok:false, disclosed in-band, not an export/receipt) and O2/R3 observations remain observations, no conditions. Fixture Herald key is a test double (pinned herald-fixture, keys/ gitignored mode 600); Promptseal consumes Herald-signed approval records, issues none, redefines none; value_class:"test" throughout.

Usage: not collected — by design

Cyan Eclipse Faircharge

Subscription-creep measurement over fixture charge streams against a sealed original authorization — creep is detected against the sealed price history from the original authorization, so a merchant's gradual increases are measured against proof, not memory. PASS (local, fixture charge streams + sealed authorizations) — local stdio MCP, fixture charge streams + sealed authorizations only, value_class "test"; Faircharge measures drift — it makes no breach, overcharge, or misconduct finding.

authorization_seal · charge_measure · relationship_reauthorize · relationship_status · creep_receipt_get · measurement_verify · measurement_pack_export · relationship_close · anomaly_queue_read

Test Bench: PASS (local, fixture charge streams + sealed authorizations) — PASS (local, fixture charge streams + sealed authorizations) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-faircharge.md) PASS-CONDITIONAL (local, fixture charge streams + sealed authorizations) with conditions C1–C7 plus spec-hygiene condition C0 (spec §2.2.2 formula sentence contradicted fixtures F-FC4/F-FC8; Verity ruled the fixture/defining-test reading governs and the shipped behaviour IS the fixture's; §2.2.2 + §6.4 step-3 amended to it); fix round v0.1.1 closed C0–C7; delta re-bench of v0.1.1 (rebench-faircharge.md) and round 2 of v0.1.2 (rebench2-faircharge.md) raised D1–D6 and E1–E4; delta re-bench round 3 of v0.1.3 (rebench3-faircharge.md): E1 (pack face binding) + E2 (anomaly gate coverage) + E3 (envelope) + E4 (whole-store validation) CLOSED, C0–C7 and D1–D6 hold under serialized re-runs and full-strength frontier sweeps — CONDITIONS CLOSED, Faircharge — PASS (local, fixture charge streams + sealed authorizations). Builder suite re-run by the bencher at every bench: 18/18 at v0.1.0, 44/44 at v0.1.3 (38 prior + 6 round-3 fixes); R3 independent probes: 213 PASS / 1 probe-bug FAIL, standalone re-run PASS = 214/214 effective serialized checks. Decisive fixture F-FC8 independently reproduced at every bench: counter-signed seal (both thresholds pinned minor-only, scheduled step 1050@idx4), charges [1000,1030,1090,1050,1090,1095,1200] → exact sequence WITHIN, WITHIN, CREEP_PER_CHARGE, WITHIN, WITHIN, CREEP_CUMULATIVE, CREEP_BOTH; serialized pack verifies in a fresh process from pack + pinned public keys alone. PASS (local, fixture charge streams + sealed authorizations) is not CERTIFIED — certified is false, and nothing beyond the PASS verdict at its exact scope is claimed. Scope is exactly: local, fixture charge streams + sealed authorizations only; no live merchant, card network, bank feed, billing system, or real charge is touched; no real consumer financial data enters any fixture; value_class:"test" throughout. SPEC §11.2 measurement-not-breach fence: Faircharge measures drift against the sealed original authorization. Whether a measured increase breaches the merchant's terms or any law is a contract/legal question for Docket/counsel — Faircharge makes no breach, overcharge, or misconduct finding. A CREEP verdict is a measurement, not an accusation and not a breach finding; WITHIN_TERMS means only that the stream measured within its own sealed thresholds (or that a dimension was unpinned, labelled on the face — §6.3 no-verdict-laundering). The SPEC §10.1 counsel gates are UNBUILT, named in not_tested. Novel mechanism (SPEC §2.1, verbatim): creep is detected against the sealed price history from the original authorization, so a merchant's gradual increases are measured against proof, not memory. Stored serves in v0.1.3 are verify-before-serve (receipts hash/signature-gated, C3), the canonical measurement pack binds all 15 face fields exactly (E1, verifier_transcript dropped), the anomaly queue gates every store-consuming surface (E2), the stdio envelope survives non-object requests (E3), and mutators whole-store-validate before any write (E4). Non-blocking hygiene note F1 stands, verbatim class: container-level foreign keys undetected + re-persisted (never served, no verdict impact) (full text in not_tested).

Usage: not collected — by design

Cyan Eclipse Timecapsule

Commitment-first track-record registry over fixture commitments — predictions and SLA commitments are sealed publicly before the outcome window and scored at maturity — a track record that cannot be backdated or cherry-picked. CERTIFIED (local/simulator, public seal registry over fixture commitments, value_class "test") — local stdio MCP, fixture commitments + fixture outcomes only, value_class "test"; no live registry, no enforceable SLAs, no reputation rating.

commitment_seal · commitment_status · author_counter_sign · track_record_get · registry_open · outcome_present · commitment_score · registry_head_get · registry_export · track_record_verify · anomaly_queue_read

Test Bench: CERTIFIED (local/simulator, public seal registry over fixture commitments, value_class "test") — CERTIFIED (local/simulator, public seal registry over fixture commitments, value_class "test") — CERTIFIED — Cyan Eclipse Timecapsule v0.1.4, local/simulator scope only: public seal registry over fixture commitments, value_class "test". No certification claimed beyond this scope. Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-timecapsule.md) PASS-CONDITIONAL (local/simulator, public seal registry over fixture commitments) with conditions C1–C8; fix rounds v0.1.1–v0.1.3 closed C1–C8, D1–D5 and N1–N3; delta re-bench round 4 of v0.1.4 (rebench4-timecapsule.md): R1 (aux anchor signature-null bypass) CLOSED, remaining conditions NONE, builder suite 69/69 re-run by the bencher, independent harness 72/72 checks PASS, regression spot probes confirm C/D classes stay closed — CERTIFIED at exactly this scope. Decisive fixture F-T10 independently reproduced at the bench: from the registry export + pinned public keys alone, an author's complete track record recomputes (8 seals, gapless author_seq, head counts exact, tallies identical) and a cherry-picked presentation of only the winners fails step 5, naming the missing author_seq values. Scope is exactly: local/simulator, public seal registry over fixture commitments, value_class "test"; fixture cryptography only (1024-bit fixture RSA, fixture keys, fixture clock); no live prediction, real SLA, real outcome feed, or public registry deployment exists in v0.1.4. Gates remain unbuilt by design and are OUT of certified scope: no public hosting, no outcome feed, no SLA enforcement, no charge path, no network. Score ≠ merit (SPEC §10.3): a track record establishes only that stated commitments were sealed before their windows and scored under their pinned rules over presented outcomes — whether the commitments were hard, the outcomes true, or the author skilled is outside what the recomputation shows. Novel mechanism (SPEC §2.1, verbatim): predictions and SLA commitments are sealed publicly before the outcome window and scored at maturity — a track record that cannot be backdated or cherry-picked. Stored serves are verify-before-serve / downgrade-labelled (C5), aux stores (metering + anomalies) are Herald-anchored with entry-by-entry validation (C6 + R1), and every anchor-consulting surface fails closed as STORE_DEFECT / AUX_ANCHOR_DEFECT on a signature-defective aux anchor with zero writes.

Usage: not collected — by design

Cyan Eclipse Burnline

Burn-trajectory forecast over sealed fixture budget mandates — recomputes the trajectory after each record and issues a receipted THROTTLE, ESCALATE, or ABORT verdict before the breach. CERTIFIED (local/simulator) — local stdio MCP, fixture spend streams + fixture Till mandates + fixture HMAC stand-ins only, pinned model burnline-linear/1.0 only; nothing certifies live budget enforcement, real Herald/Meter/Till interop, or that a verdict stops spend.

stream_open · spend_append · stream_status · intervention_get · stream_close · forecast_verify · pack_get

Test Bench: CERTIFIED (local/simulator) — CERTIFIED (local/simulator) — CERTIFIED — Cyan Eclipse Burnline v0.1.5, local/simulator scope only: fixture spend streams, fixture Till mandates, fixture HMAC stand-in signatures, pinned model burnline-linear/1.0 only. No certification claimed beyond this scope. Final verdict wording (Verity delta re-bench round 5, rebench5-burnline.md, verbatim): CERTIFIED (local/simulator) — v0.1.5. C1+C2 closed in v0.1.1, C3 in v0.1.2, C4 in v0.1.3, C5+C6+O2 in v0.1.4, C7+O3 in v0.1.5; no remaining conditions. Scope: fixture spend streams, fixture Till mandates, fixture HMAC stand-in signatures, pinned model burnline-linear/1.0 only. Nothing certifies live budget enforcement (§10 counsel gate, unbuilt), real Herald/Meter/Till interop, or that a verdict stops spend. Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave2/bench-burnline.md) PASS-CONDITIONAL with conditions C1–C7; fix rounds v0.1.1–v0.1.5 closed C1–C7 + O2 + O3; delta re-bench round 5 of v0.1.5: C7 CLOSED, O3 CLOSED, C1–C6 not reopened, remaining conditions NONE, builder suite 53/53 re-run by the bencher, independent probes 113 assertions — CERTIFIED at exactly this scope. Decisive fixture F-BU14 independently reproduced at every bench (fresh process, empty data dir): exported ladder pack verifies true and the ABORT verdict sits at exactly the first head where the pinned burnline-linear/1.0 projection crossed the abort threshold, cumulative burn 3100 < cap 10000, byte-identical on rerun. Scope is exactly: local/simulator, fixture spend streams + sealed fixture Till mandates only, fixture HMAC stand-in signatures, pinned model burnline-linear/1.0 only; value_class "test" on every artifact; exact rational arithmetic, ceiling rounding, no floats in signed objects. Burnline issues verdicts only — it holds no spend authority, blocks no transaction itself, and moves no money; nothing certifies that a verdict stops spend (enforcement is the consuming runtime's act, and a runtime that ignores a verdict is recorded as VERDICT_UNACKNOWLEDGED). Two limits stated wherever the mechanism is described: (a) intervention is possible only when the trajectory crosses a threshold before the cap — a single record can jump the cap with no prior intervention possible (F-BU8, NO_PRIOR_INTERVENTION_POSSIBLE); (b) the projection is the pinned model's arithmetic over the records presented, not a prediction about any real system's future spend; the model is never a learned model. Novel mechanism (SPEC §2.1, verbatim): burn-trajectory forecasting over sealed budget mandates intervenes before breach — throttle, escalate, or abort — and the intervention itself is receipted. Stored serves and every mint path verify before serving/minting (C1–C7 closures): chain, burn, receipts, acks and the recomputed defect set are recomputed from source data; store containers (root, stream, metering, refusal entries) carry exact key sets — foreign keys are detected and refused, never silently re-persisted (O3); a tampered store yields a named BURNLINE_VERIFY_FAIL refusal or a downgraded pack, never the stored assertion and never a signed close over unverified bytes.

Usage: not collected — by design

Cyan Eclipse Shelfspace

Sponsorship-disclosure witness for agent-facing marketplaces — paid placements carry a sealed sponsorship receipt, and a paid placement with no valid receipt derives UNMARKED-PAID. PASS-CONDITIONAL (local, fixture placement records) — local single-operator file model, fixture placements only. Shelfspace issues sponsorship receipts and a surfacing-proof format and verifies proofs presented to it. It does not compel any router, marketplace, or agent to surface a disclosure. A 'required to surface' duty exists only where a router/marketplace adopts it (§12); Shelfspace evidences surfacing, it does not enforce it. Honesty fence — rollback residual: within this local file model, a rollback that restores every local artifact (store, both anchors, journal, metering) to one consistent earlier snapshot is undetectable; the journal/metering cross-checks detect partial rollbacks and single-artifact tampering, not a complete coordinated snapshot restore. No stronger claim is made.

placement_record · sponsorship_receipt_issue · sponsorship_receipt_revoke · placement_status · shelfspace_pack_get · render_record · surfacing_proof_issue · surfacing_proof_get · receipt_verify

Test Bench: PASS-CONDITIONAL (local, fixture placement records) — PASS-CONDITIONAL (local, fixture placement records) — Verity delta re-bench 2 of v0.1.2, 2026-10-27 (argus101-program-20261027/wave2/rebench2-shelfspace.md): all six re-bench conditions F1'–F6' CLOSED; F2' snapshot-rollback residual adjudicated acceptable-at-scope under condition 1; closed round-1 items (cond.2/3/4, F-H14) spot-checked, none reopened; builder suite 52/52 re-run by the bencher plus ~200 independent serialized probes. PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond the conditional verdict at its exact scope is claimed. Scope is exactly: local, fixture placement records, under the local single-operator file model; local stdio MCP, fixture placements only, value_class "test" on all artifacts, fixture HMAC stand-in for the Herald pattern, no counsel-gate (§10) surface built or claimed. CONDITION 1 (the sole condition, carried verbatim): Scope/fence condition (from F2' residual): certification holds only at scope local, fixture placement records under the local single-operator file model. The README rollback-residual honesty fence and scope label must remain on the product face. Any live, multi-party, shared-filesystem, or adversarial-operator use requires an anchor outside the data-dir trust model + fresh bench (the existing counsel gate for live use stands separately). Decisive fixture F-H14 stands from the bench record: plc-paid-1 SURFACED with proof buying_agent_ref/disclosure_marker bound, plc-paid-2 derives UNMARKED-PAID / UNMARKED_PAID_NO_RECEIPT, fresh verify PASS with eval bound to the signed statement.

Usage: not collected — by design

Cyan Eclipse Safelist

Freshness-windowed signed allowlist feed for agent services — feed entries expire unless re-benched inside the window; the allowlist is a live verdict stream, not a static directory. CERTIFIED (local, fixture feed entries) — local stdio, fixture services/receipts only, value_class "test". A Safelist entry carries a bench verdict exactly as its receipt states it, scope qualifier attached. ACTIVE means only: a verdict receipt of that stated scope was verified, and the entry is inside its freshness window at the stated evaluation time. It does not mean the service is safe, endorsed, compliant, or currently passing anything.

safelist_feed_open · safelist_entry_add · safelist_entry_rebench · safelist_verdict_revocation_present · safelist_entry_withdraw · safelist_feed_pack_get · safelist_feed_get · safelist_entry_status · safelist_feed_verify

Test Bench: CERTIFIED (local, fixture feed entries) — CERTIFIED (local, fixture feed entries) — Verity spot re-bench 3 of v0.1.3, 2026-10-27 (argus101-program-20261027/wave2/rebench3-safelist.md): R4-D2 CLOSED, conditions NONE; builder suite 50/50 re-run twice by the bencher, independent probes 23/23 + 93/93 behavioral (14-field transcript sweep) + 11/11 (verdict/transcript mismatch edges + regression spot). Scope is exactly: local, fixture feed entries — local stdio, fixture services/receipts only, value_class "test" throughout, no live listing/routing/reliance/bench-sale path; counsel gate (§10) unbuilt by design. Decisive fixture F-L13 stands from the bench record: pack alone ACTIVE@T0+3599, EXPIRED@T0+3600, EXPIRED@T0+7200; backdated receipt (bench_ts=T0+3500, presented now=T0+7200) → REBENCH_LATE.

Usage: not collected — by design

Cyan Eclipse Muzzle

Cyan Eclipse Muzzle — PASS (local/fixture) — Verity reduced-scope bench, secrets and PII only, ruleset muzzle-rules/0.4. A PASS verdict is scan evidence, not a guarantee of safety or non-defamation.

muzzle_scan · muzzle_rescan_status · muzzle_receipt_get · muzzle_receipt_verify · muzzle_require_pass · muzzle_export

Test Bench: PASS-CONDITIONAL (local/fixture) — PASS-CONDITIONAL (local/fixture) — Verity reduced-scope bench of v0.1.3, ruleset muzzle-rules/0.4, 2026-10-27 (argus101-program-20261027/ruling-muzzle-reduced-scope.md), secrets and PII limbs ONLY: fresh bench-authored corpus (36 items, disjoint in values/shapes/phrasing from all repo fixtures and both prior Verity corpora), all rates computed from signed receipts after every receipt was independently re-verified — secrets 100% recall / 100% precision (13/13, 0 FP), PII 100% recall / 100% precision (13/13, 0 FP; spec thresholds ≥95% / ≥90%), benign controls 10/10 unflagged, verdict agreement 36/36; tamper battery, redaction invariant (zero raw-value escapes over corpus + canary sweeps), downgrade law, fail-closed fence (7/7), and exact 6-tool surface all PASS. PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond the conditional verdict at its exact scope is claimed. Sole condition: the ruling §9 authorised claim block ships verbatim on every claim surface (README, this listing, server.legacy.json full manifest); the registry server.json description is the ≤100-char registry short form carrying the §11.1 sentence verbatim and claiming only secrets + PII. This release does not scan for defamation-risk: defamation-risk detection is excluded from the certified scope of this version, and a PASS verdict is not evidence about defamation. Defamation detection is not included in v0.1 — parked for a v0.2 redesign.

Usage: not collected — by design

Cyan Eclipse Sandclock

Eval-expiry clock over fixture Benchkit-format scorecards + fixture drift evidence — every eval score carries an expiry computed from measured model/tool drift rates; consumers get a 'stale eval' verdict when a score outlives its evidence. PASS-CONDITIONAL (local) — local stdio, fixture scorecards + fixture drift evidence only, value_class "test"; no live model/tool observation, no eval is run, no score is issued. Expiry is evidence-decay arithmetic, not a claim the score was wrong.

expiry_issue · expiry_status · stale_check · expiry_verify · expiry_export

Test Bench: PASS-CONDITIONAL (local) — PASS-CONDITIONAL (local) — Verity delta re-bench 2 of v0.1.2, 2026-10-27 (argus101-program-20261027/wave2/rebench2-sandclock.md): F1–F4 closed (store set-anchor / key-binding / mint-signing / envelope findings), F1 whole-file residual adjudicated acceptable-at-scope under condition 1; builder suite 50/50, independent probes 55/55; closed F1–F6 first-pass items spot-checked — none reopened. PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond the conditional verdict at its exact scope is claimed. Scope is exactly: local (fixture scorecards + fixture drift evidence only, value_class "test") under the local single-file / single-operator store model — local stdio MCP (SPEC §0.1); no live model/tool observation, no eval run, no score issued; the counsel-gated live phase (SPEC §10.2) remains unbuilt and separately gated. Condition 1 (verbatim): certification holds only at scope local (fixture scorecards + fixture drift evidence only, value_class:"test") under the local single-file / single-operator store model. The README whole-file-rollback residual fence + scope label must remain on the product face. Any live, multi-party, shared-filesystem, or adversarial-operator use requires an anchor outside store.json (e.g. receipt-held head hash / external append-only log) + fresh bench; the counsel-gated live phase (spec §10.2) remains unbuilt and separately gated.

Usage: not collected — by design

Cyan Eclipse Archivist

Web-evidence capture witness over fixture captures — page captures seal content, headers, and the TLS certificate together as one hash-chained, Herald-signed bundle proving what was served, by whom, and when. PASS-CONDITIONAL (local, fixture captures) — local stdio, fixture captures only, value_class "test"; no live web capture exists in v0.1. This bundle proves what the capturing fetcher was served, at the witness time, under the certificate presented to it. A server may serve different content to other fetchers, at other times, or under other certificates — the bundle says nothing about what anyone else was served.

capture_present · bundle_seal · bundle_status · bundle_get · bundle_verify · bundle_pack_export · anomaly_queue_read

Test Bench: PASS-CONDITIONAL (local, fixture captures) — PASS-CONDITIONAL (local, fixture captures) — Verity delta re-bench 2 of v0.1.2, 2026-10-27 (argus101-program-20261027/wave2/rebench2-archivist.md): R1–R8 all closed (cross-artifact store head / signed bundle-set anchor / sealed metering / record_id binding / exact pack face / structured whole-store garbage / missing-key zero-write / sealed-snapshot defect findings), no closed F1–F6 item reopened; builder suite 36/36 plain + socket-denied, independent probes 135 serialized assertions; the one residual (full coordinated snapshot restore) adjudicated a documented condition, fence present and accurate on README + every pack + server.json. PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond the conditional verdict at its exact scope is claimed. Scope is exactly: local, fixture captures only (fixture page content, fixture headers, fixture TLS material, value_class "test") under the local single-file / single-operator store model — local stdio MCP (SPEC §0.1); no live web capture exists in v0.1: no fetcher, no network client, no real page, certificate, or server is touched. Condition 1 (verbatim): Rollback residual (R1c): a full coordinated restore of all local artifacts to one consistent earlier snapshot is undetectable in the local file model — fenced verbatim on README + every pack (rollback_fence) + server.json. External anchor required before any live or multi-party use/claim. Condition 2 (verbatim): §10 counsel-gated phases remain unbuilt and unclaimed; any claim at those scopes requires counsel clearance + scope edit + re-bench (§10.2).

Usage: not collected — by design

Cyan Eclipse Allowance

Sealed family budget mandates for household agents — category caps, merchant lists, and amounts are sealed by the account holder; every spend lands in a family ledger as a receipt the agent cannot edit. PASS-CONDITIONAL (local/simulator) — local stdio, fixture family mandates + fixture spends only, value_class "test"; no real money, no payment rail, counsel gate (§10) unbuilt. Ledger receipts are Herald-counter-signed; the household agent holds no ledger signing key.

mandate_seal · mandate_amend · mandate_revoke · ledger_get · ledger_verify · refusal_log_get · spend_present · mandate_status · spend_history_get

Test Bench: PASS-CONDITIONAL (local/simulator) — PASS-CONDITIONAL (local/simulator) — Verity delta re-bench 2 of v0.1.2, 2026-10-27 (argus101-program-20261027/wave2/rebench2-allowance.md): R1–R7 all CLOSED on fresh repros (per-mandate chains + Tier-M/Tier-G isolation; sealed metering; fail-closed holder-signature verify; exact seal field sets; monotonic amendments; structured store-defect faces with zero raw exceptions; subset restores detected), R7 full-restore residual adjudicated as fenced (condition, not fail), regression spot PASS (F-A13 decisive fresh-process recompute; F8 §7.4 split key custody re-verified); builder suite 44/44 PASS x2 clean runs, independent probes ~110 checks. PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond the conditional verdict at its exact scope is claimed. Scope is exactly: local/simulator only, fixture family mandates only (fixture holders, members, agents, merchants, categories and spends, value_class "test"), local stdio MCP (SPEC §0.1); no real money, no payment rail, no merchant, bank, or household account is connected in v0.1.2, and Allowance moves no money even in simulation — it checks and records fixture spends. Condition 1 (verbatim): R8 — cross-mandate transplant id↔hash pairing is detected only indirectly (transplanted record never served/counted/exported); Tier-G structural check or README fence stands as the condition. Bench condition text (verbatim): R8: add the id↔hash pairing to the structural/Tier-G check (an entry whose `mandate_id` and `mandate_hash` name different mandates defects the store on every path, incl. target-mandate spend), or fence the pairing class in the README integrity section. Condition 2 (verbatim): O1 — ledger_get VERIFIED slice-export vs downgraded faces on a Tier-G-defected store (documented slice semantics; no false data). Bench condition text (verbatim): O1: make `ledger_get` downgrade when Tier-G defect exists outside the served slice, or document the slice-only semantics on the export face.

Usage: not collected — by design

Cyan Eclipse Receiptvault

Purchase-receipt vault with a gaslight-proof return countdown — return deadlines are computed at purchase time from the merchant's sealed terms, so the countdown can't be gaslit later; alerts fire before windows close. CERTIFIED (local, fixture sealed merchant terms + fixture purchase records) — local stdio, fixture sealed merchant terms + fixture purchase records only, value_class "test"; the deadline is arithmetic over the merchant's terms as sealed at purchase; it is not a legal determination, and Receiptvault does not warrant that any merchant will accept a return by it. Standing condition: full coordinated local snapshot restore is undetectable in the local model — fenced verbatim in README + server.json; an external anchor (witnessed/countersigned head hash outside the local store) is REQUIRED before any live or multi-party use.

receiptvault_open · receiptvault_status · receiptvault_mark_returned · receiptvault_get · receiptvault_verify · receiptvault_list · receiptvault_clock_advance · receiptvault_anomaly_queue_read

Test Bench: CERTIFIED (local, fixture sealed merchant terms + fixture purchase records) — CERTIFIED (local, fixture sealed merchant terms + fixture purchase records) — Verity delta re-probe (re-bench 3) of v0.1.3, 2026-10-27 (argus101-program-20261027/wave2/rebench3-receiptvault.md): RB-F9 CLOSED on fresh repros (dangling/repointed purchase index and deleted-field entries defect as SEAL_DEFECT on every surface, no raw KeyError anywhere, stdio process survives, zero writes on defect paths), RB-F1–RB-F8 closed at re-bench 2 and not reopened, regression spot PASS (F-RV9 decisive fresh-process recompute holds; full-export head-anchor/record-count drops fail verification); builder suite 66/66 PASS plus socket-denied 66/66, bencher battery 130/130 PASS. Scope is exactly: local, fixture sealed merchant terms + fixture purchase records — local stdio MCP (SPEC §0.1), fixture purchases, merchants, terms, seals, holiday calendars, clock and alert sink only, every artifact value_class "test"; no live merchant, storefront, receipt inbox, terms page, or notification channel is touched in v0.1.3 and no real purchase or terms are vaulted. Standing condition (verbatim): full coordinated local snapshot restore is undetectable in the local model — fenced verbatim in README + server.json; an external anchor (witnessed/countersigned head hash outside the local store) is REQUIRED before any live or multi-party use.

Usage: not collected — by design

Cyan Eclipse Scamshield

First-payment friction layer for agent-assisted payments — a first payment to any new payee triggers a witnessed risk briefing (fraud reports, payee-continuity, lookalike domains) the user must acknowledge; the acknowledgment is receipted. PASS-CONDITIONAL (local/simulator, fixture risk sources + fixture payee history) — local stdio, fixture risk sources + fixture payee history only, value_class "test"; Scamshield assembles a briefing from witnessed sources and receipts the user's acknowledgment of that briefing; it does not certify a payee safe, trustworthy, legitimate, or fraudulent, and it issues no risk score or verdict on any payee. Condition 1: RB-F7 coordinated-restore residual stays fenced verbatim on README + all named faces — full coordinated restore of all artifacts (incl. store_head.json) is undetectable in the pure local model; an external anchor is required for detection. Condition 2: scope-only certification — the live phase remains behind the §10 counsel gate + separate re-bench.

scamshield_check · scamshield_briefing_get · scamshield_acknowledge · scamshield_status · scamshield_verify · scamshield_list · scamshield_anomaly_queue_read

Test Bench: PASS-CONDITIONAL (local/simulator, fixture risk sources + fixture payee history) — PASS-CONDITIONAL (local/simulator, fixture risk sources + fixture payee history) — Verity delta re-bench 2 of v0.1.2, 2026-10-27 (argus101-program-20261027/wave2/rebench2-scamshield.md): RB-F1–RB-F8 all CLOSED on fresh repros (genuine-export anchor/count binding; key preflight with byte-identical disk on KEY_MISSING; structured stdio envelope; exact nested field sets with no raw exceptions; source-intent-history injection downgrades at the exact named verify step; temporal-order gates; subset restores detected via the retained store_head.json cross-artifact commitment; duplicate sources refused), RB-F7 coordinated-restore residual adjudicated as a fenced CONDITION (not a fail), regression spot PASS (F-SS9 decisive fresh-process recompute; ack same-briefing binding; closed originals not reopened); builder suite 53/53 PASS x3 by the bencher, ~120 independent probe checks. PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond the conditional verdict at its exact scope is claimed. Scope is exactly: local/simulator, fixture risk sources + fixture payee history — local stdio MCP (SPEC §0.1), fixture payment intents, payees, histories, fraud reports, continuity chains, lookalike findings, users and keys only, every artifact value_class "test"; no live payment, payee, fraud database, domain registry, or user is touched in v0.1.2, and Scamshield scores, blocks, releases, or executes no payment and certifies no payee. Condition 1 (verbatim): RB-F7 residual (fenced): a full coordinated restore of every local artifact (store.json, store_head.json, and any backups) is undetectable in the pure local model and is NOT claimed detected; detection requires an external anchor outside this local store. The fence must remain verbatim on README + status/list/anomaly/verify/briefing_get/check faces in any future version claiming this certification. Condition 2 (verbatim): Certification extends only to the scope above; any live phase remains behind the §10 counsel gate and requires a separate re-benched build.

Usage: not collected — by design

Cyan Eclipse Holdfast

Simulated escrow-release witness over a fixture escrow ledger — release is triggered by a counter-signed acceptance proof generated by executing the buyer's own acceptance tests — the test-run receipt, not a human click, unlocks funds. PASS-CONDITIONAL(simulator, fixture escrow ledger) — fixture escrow ledger only; no real funds are held, moved, or owed. Condition C-1: external anchor required before live/multi-party use — coordinated rollback across artifacts is not claimed by this local build. Condition C-2: SETTLEMENT_BPS = 50 requires price-book/owner ratification before priced reliance (SPEC states no numeric figure — §9.1 shape only, §9.2 defers to price book; the 50 bps component is disclosed and NEVER collected in this scope, collected:false on every metering record).

holdfast_terms_propose · holdfast_terms_sign · holdfast_fund_sim · holdfast_delivery_present · holdfast_proof_present · holdfast_status · holdfast_verify · holdfast_dispute_handoff · holdfast_escrow_list · holdfast_anomaly_queue_read

Test Bench: PASS-CONDITIONAL(simulator, fixture escrow ledger) — PASS-CONDITIONAL(simulator, fixture escrow ledger) — Verity bench of v0.1.0, 2026-10-27 (argus101-program-20261027/wave3/bench-holdfast.md): builder suite 34/34 re-run clean by bencher; decisive fixture F-H9 independently re-derived in both runs from bencher corpus + bencher keys; bencher adversarial battery 60/60 (recompute-in-verify, signed head + SET anchors, signed-records-only derived state, exact face/field sets, whole-store validation pre-write, consume-or-refuse partition with byte-identical store on every refusal, counter-signatures on every path, fail-closed verify on missing key, temporal monotonicity, crash-proof envelopes, F-H2/F-H5/F-H6 state battery, F-H3 expiry return, amount discipline, no-click law with 13 forbidden verbs absent, dispute handoff, labels on 100% of artifacts). Defects: none. PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond the conditional verdict at its exact scope is claimed. Scope is exactly: local stdio MCP, fixture escrow ledger only, value_class "test" on every artifact and simulated: true on every ledger artifact; no real funds are held, moved, released, or owed, and no custody in any form exists in v0.1.0. Condition C-1: external anchor required before live/multi-party use — coordinated rollback across artifacts is not claimed by this local build. Condition C-2: SETTLEMENT_BPS = 50 requires price-book/owner ratification before priced reliance (SPEC states no numeric figure — §9.1 shape only, §9.2 defers to price book; the 50 bps component is disclosed and NEVER collected in this scope, collected:false on every metering record).

Usage: not collected — by design

Per-call voice continuity tokens over fixture call records — per-call continuity tokens seal the speaker model + voice fingerprint, so a voice agent cannot silently swap voices mid-relationship. PASS-CONDITIONAL(local, fixture call records) — fixture call records only; the claim is provable only at identifier granularity and same-model re-voicing behind unchanged identifiers is not detectable; never 'same voice verified', no biometric identity.

relationship_enrol · call_seal · relationship_rebaseline · relationship_close · relationship_status · continuity_pack_get · chain_verify · token_verify

Test Bench: PASS-CONDITIONAL(local, fixture call records) — PASS-CONDITIONAL(local, fixture call records) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave3/bench-voiceprint.md): builder suite 32/32 re-run clean by bencher; decisive fixture F-V14 independently reproduced on all three branches from the exported pack + pinned keys alone in a fresh process; bencher battery 82/82 after 3 probe-artifact corrections (F-1) — swap/drift exactness (drift never returned as SWAP_DETECTED; binary-exact 0.25 at threshold CONTINUOUS, 0.375 SUSPECT), claim-surface audit (narrowed claim verbatim on README/server.json/statements/verify results; forbidden §11.4 strings zero hits outside SPEC's own enumeration), 14-point delta law, refusal/gap battery, tamper battery (downgrade-never-upgrade on every path), metering exactness, disjoint surfaces, §7.3 fence, privacy fence (no audio capture/playback, telephony, speech, or biometric code path in the tree). Sole condition C1 (verbatim): external anchor required before live/multi-party use (cross-artifact coordinated rollback fenced; local SET anchor binds head + record_count within one pack only). Advisories F-2/F-3/F-4 (non-blocking, disclosed): fixture feature vectors are retained in attestation objects carried in the pack (token artifacts store the hash only; accepted at fixture scope, re-bench required for any real feature data); the §11.2 legend's '(§3.5)' parenthetical appears in README only; token_verify accepts a sealed token's SUSPECT_FEATURE_DRIFT on its identifier aspect only (full drift recompute lives in chain_verify). Scope is exactly: local stdio MCP, fixture call-record metadata only, value_class 'test' on all artifacts; no real audio is captured, processed, or stored, no human speaker is identified, and no live call platform is touched. Novel mechanism (SPEC §2.1, verbatim): per-call continuity tokens seal the speaker model + voice fingerprint, so a voice agent cannot silently swap voices mid-relationship.. Narrowed claim (verbatim): The 'cannot silently swap' claim is provable only at identifier granularity: a swap is proven only by a change in the attested model_id/voice_id (or a chain break); same-model re-voicing behind unchanged identifiers is not detectable.

Usage: not collected — by design

Cyan Eclipse Understudy

Delegation subset proofs over Till-grammar fixture mandate pairs — a cryptographic subset proof that the sub-agent's mandate is a strict attenuation of the delegator's. PASS-CONDITIONAL(local, subset proof over Till-grammar fixture mandates) — fixture mandate pairs only; issues no mandates, grants no authority, registers no agents.

understudy_prove · understudy_proof_get · understudy_verify · understudy_chain_verify · understudy_evaluation_list · understudy_anomaly_queue_read

Test Bench: PASS-CONDITIONAL(local, subset proof over Till-grammar fixture mandates) — PASS-CONDITIONAL(local, subset proof over Till-grammar fixture mandates) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave3/bench-understudy.md): builder suite 34/34 re-run clean by bencher; decisive fixture F-U9 independently reproduced (bencher corpus, bencher keys, black-box stdio) — STRICT_ATTENUATION with face vector and subset_proof equal to bencher recomputation, full chain recomputed from the §6.2 export alone, and fresh-process empty-data-dir verify → VERIFIED_EXPORT / STRICT_ATTENUATION re-derived; all 8 F-U5 single-dimension widenings NOT_SUBSET with named codes and fresh-process recompute; 9-field absence totality, §5.3 refusal partition, verify/tamper/downgrade battery, metering exactness, and malformed battery all pass (bencher battery 99 checks). PASS-CONDITIONAL is not CERTIFIED — certified is false. Scope is exactly: local stdio, fixture mandate pairs only, value_class 'test'; Understudy issues no mandates, grants no authority, registers no agents, and executes no delegated action. Sole condition (verbatim): Cross-artifact cross-commitment note: full coordinated rollback across separately held artifacts is a fenced condition — external anchor required before live/multi-party use. Advisory F-2/F-3 (non-blocking, v0.2 hardening notes): chain-hop hash continuity (chain_verify binds hops by mandate_id, not hash/citation) and offline verify temporal re-check (verify step 2 does not re-run write-time predates/expiry-at-issuance checks; prove, the sole mint path, refuses both).

Usage: not collected — by design

Agent birth-certificate issuer — at registration the agent's creator, model, config hash, and declared purpose scope are sealed; acting outside the declared purpose is detectable by any counterparty who checks the certificate. PASS-CONDITIONAL(local/fixture) — local stdio, fixture agents/creators/keys only, value_class "test"; Genesis detects, it never blocks. Sole condition: full coordinated rollback is a fenced condition — external anchor required before live/multi-party use.

genesis_certificate_register · genesis_certificate_amend · genesis_certificate_revoke · genesis_certificate_get · genesis_certificate_verify · genesis_purpose_detect · genesis_certificate_export · genesis_family_list · genesis_metering_read

Test Bench: PASS-CONDITIONAL(local/fixture) — PASS-CONDITIONAL(local/fixture) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave3/bench-genesis.md): builder suite 43/43 re-run clean by the bencher; decisive fixture F-G13 independently re-derived from the bencher's own corpus and bencher-generated keys — registration agent_ref matched the bencher-predicted mint, exported certificate / seed prev_hash / creator + witness signatures / export head anchor all recompute under bencher code, fresh-process verify from export + pinned keys alone returned SEAL_VALID, and detection over the bencher's 5 actions reproduced exactly [IN_PURPOSE, IN_PURPOSE, IN_PURPOSE, OUT_OF_PURPOSE, INDETERMINATE] with the out-of-purpose record naming clause action_types and the indeterminate record naming reason unmapped, never coerced. PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond this scope is claimed. Scope is exactly: local stdio, fixture agents/creators/keys, fixture Docket/Meter-grammar action records, value_class "test"; no live registry, no third-party reliance, no enforcement use. Sole condition (verbatim): cross-artifact cross-commitment — full coordinated rollback = fenced condition — external anchor required before live/multi-party use. Genesis detects, it never blocks: no blocking/enforcement verb exists on either surface (dispatch enumeration by the bencher).

Usage: not collected — by design

Schedule-integrity guard for recurring agent payments — any witnessed counterparty behavior change (price drift, endpoint change, ownership change) automatically freezes the schedule until re-attestation. PASS-CONDITIONAL(simulator, fixture schedules + witnessed-change feed) — fixture schedules + fixture witnessed-change feed only; no payments are executed, initiated, or held. Condition C-1: cross-artifact cross-commitment is the local per-schedule hash chain + export SET anchors only; full coordinated rollback not provided — external anchor required before live/multi-party use.

slowpay_schedule_create · slowpay_baseline_seal · slowpay_due_simulate · slowpay_reattest · slowpay_schedule_cancel · slowpay_status · slowpay_verify · slowpay_change_present · slowpay_schedule_list · slowpay_anomaly_queue_read

Test Bench: PASS-CONDITIONAL(simulator, fixture schedules + witnessed-change feed) — PASS-CONDITIONAL(simulator, fixture schedules + witnessed-change feed) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave3/bench-slowpay.md): builder suite 42/42 re-run clean by bencher; bencher black-box suite 82/82; decisive fixture F-S9 independently reproduced in both runs (genuine witnessed price drift 4321→4400 freezes in the presenting evaluation, held due event never back-processed, fresh-process verify from export + pinned keys alone returns verified: true / ACTIVE after buyer re-attestation; the one-character altered copy 4400→4409 returns RECEIPT_HASH_MISMATCH, no freeze record, schedule remains ACTIVE). No defects found. Sole condition C-1: cross-artifact cross-commitment is the local per-schedule hash chain + export SET anchors only; full coordinated rollback not provided — external anchor required before live/multi-party use (bench text: cross-artifact cross-commitment in v0.1 is the local per-schedule hash chain plus the export's SET anchors only; full coordinated rollback across artifacts is not provided — external anchor required before live/multi-party use). No live or multi-party use is licensed until that anchor exists and the build is re-benched. PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond the conditional verdict at its exact scope is claimed. Scope is exactly: local stdio MCP, fixture schedules + fixture witnessed-change feed only, value_class 'test' on all artifacts, due records simulated notes (simulated: true); no payment is executed, initiated, or held, and no live feed, mandate issuance, or witness product is touched. Novel mechanism (SPEC §2.1, verbatim): any witnessed counterparty behavior change (price drift, endpoint change, ownership change) automatically freezes the schedule until re-attestation. Advisories A-1/A-2 (non-blocking bench notes, do not limit the verdict at the stated scope): A-1 — status/list EXPIRED overlay also applies to a DRAFT schedule past mandate expiry (display-only; no transition or write; sealing against an expired mandate is refused); A-2 — in a store already containing Herald-signed freeze records, removing the Herald key surfaces as CHAIN_BROKEN (whole-store validation) rather than the mint-specific error, with the isolated clean-store probe confirming the mint gate itself (structured error, zero writes, schedule stays ACTIVE). Signing is an HMAC fixture stand-in for Herald RS256 (disclosed on every signed face as 'herald-fixture (fixture HMAC stand-in)'); a live phase would need real Herald signing and a re-bench.

Usage: not collected — by design

Patient-buying mandate over fixture balance + fixture witnessed prices — the mandate self-executes only when two witnessed conditions coincide — goal balance reached and a witnessed price at or below the target — with progress receipts throughout. PASS-CONDITIONAL(simulator) — simulator only; execution is a mandate state transition only, NO purchase is executed and no funds move. Sole condition: cross-artifact cross-commitment is fenced — external anchor required before live/multi-party use.

mandate_create · balance_present · price_present · mandate_evaluate · mandate_status · progress_get · execution_get · mandate_cancel · mandate_list · execution_verify · dispute_pack_get · anomaly_queue_read

Test Bench: PASS-CONDITIONAL(simulator) — PASS-CONDITIONAL(simulator) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave3/bench-layaway.md): builder suite 33/33 re-run clean by bencher; decisive fixture F-L13 re-derived from the bencher's own corpus (bencher battery 90/90) — execution at exactly the coincidence tick and no other, both witnesses recomputed fresh-process two ways. PASS-CONDITIONAL is not CERTIFIED — certified is false. Scope is exactly: local stdio, fixture goal-balance ledger + fixture witnessed prices, value_class 'test'; execution = a mandate state transition only — nothing was bought, ordered, or paid. Sole condition (verbatim): cross-artifact cross-commitment is fenced — external anchor required before live/multi-party use. Cross-artifact cross-commitment (mandate chain + balance segment + price observation bound by hash citations in one execution record) is a fenced condition — external anchor required before live/multi-party use Advisory A-1/A-2 (non-blocking): the config fence scans all environment values fail-closed, and signing is an HMAC fixture stand-in for Herald RS256 (disclosed; live phase would need real Herald signing + re-bench).

Usage: not collected — by design

Cyan Eclipse Duet

Dual-control attestation for fixture agent actions — two independent agents must counter-sign, and the pair must prove model diversity (different model families). PASS-CONDITIONAL(local, attested model-family declarations) — fixture proposals, fixture agents and fixture identity records only; Duet executes, authorizes, and approves nothing. Duet's diversity/independence proofs establish attested model families/providers and bound signatures — NOT actual running models, NOT non-collusion. Sole condition: cross-artifact cross-commitment is fenced — external anchor required before live/multi-party use.

duet_propose · duet_sign · duet_status · duet_attestation_get · duet_verify · duet_proposal_list · duet_anomaly_queue_read

Test Bench: PASS-CONDITIONAL(local, attested model-family declarations) — PASS-CONDITIONAL(local, attested model-family declarations) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave3/bench-duet.md) + D1 spot re-bench (wave3/rebench-duet.md, D1 CLOSED): builder suite 31/31 re-run clean by bencher at bench, 32/32 after the D1 fix (the +1 is the D1 regression test); decisive fixture F-D9 independently re-derived (bencher corpus, bencher keys, black-box stdio) — seal via stdio → SEALED with face families [orion, vega], fresh process with empty data dir from export + pinned keys alone → verified: true / SEALED, and the same-family control (signer B's identity record family edited to orion, attestation hash recomputed, record re-sealed under the bencher issuer key, refs re-derived) fails at DUET_VERIFY_FAIL:step2_check4_distinctness, UNVERIFIED_EXPORT, attestation_claim: null — the diversity leg is load-bearing; bencher battery 72 checks, 71 PASS at bench (sole failure = D1, closed on re-bench). PASS-CONDITIONAL is not CERTIFIED — certified is false. Scope is exactly: local stdio, fixture proposals, fixture agents and fixture identity records only, value_class 'test'; attestations only — Duet executes, authorizes, and approves nothing. Sole condition (verbatim): cross-artifact cross-commitment is fenced — external anchor required before live/multi-party use. Duet's diversity/independence proofs establish attested model families/providers and bound signatures — NOT actual running models, NOT non-collusion. Advisory O1/O2 (non-blocking): pairs carrying different family taxonomies seal (face surfaces each ref's own taxonomy; §2.4 includes no taxonomy-equality check — recorded for the live-identity phase), and signing is an HMAC fixture stand-in for Herald RS256 (disclosed).

Usage: not collected — by design

Cyan Eclipse Crosspay

Pre-flight FX guard for agent cross-currency payments — rate source, rate, and spread are sealed at execution, and settlement is refused pre-flight when the spread exceeds the mandate's cap. PASS-CONDITIONAL(local/simulator, fixture FX records) — fixture FX records only; never converts, settles, or moves currency. Sole condition: cross-artifact cross-commitment is fenced — external anchor required before live/multi-party use.

crosspay_check · crosspay_verdict_get · crosspay_verify · crosspay_handoff_mark · crosspay_check_list · crosspay_anomaly_queue_read

Test Bench: PASS-CONDITIONAL(local/simulator, fixture FX records) — PASS-CONDITIONAL(local/simulator, fixture FX records) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave3/bench-crosspay.md) + delta re-bench (argus101-program-20261027/wave3/rebench-crosspay.md, D1/D2/D3 CLOSED, C2/C3/C4 discharged): builder suite 39/39 re-run clean by bencher (34 baseline + 5 fix regressions); decisive fixture F-X9 independently re-derived from the bencher's own corpus (GBP/USD pair the builder never used) — under-cap CLEAR at spread 40 bps vs cap 100 with fx_seal == bencher-recomputed seal, over-cap twin REFUSED / SPREAD_CAP_EXCEEDED at 101 bps, both exports fresh-verify in a fresh process from export + pinned keys alone. PASS-CONDITIONAL is not CERTIFIED — certified is false, and nothing beyond this scope is claimed. Scope is exactly: local stdio, fixture FX records only, pre-flight verdicts only, value_class 'test'; Crosspay never converts, settles, or moves currency and fetches no live rates. Sole remaining condition (verbatim): cross-artifact cross-commitment is fenced — external anchor required before live/multi-party use. Novel mechanism (SPEC §2.1, verbatim): rate source, rate, and spread are sealed at execution, and settlement is refused pre-flight when the spread exceeds the mandate's cap — enforcement before money moves, not evidence after.

Usage: not collected — by design

Cyan Eclipse Phantom

Phantom-charge sentinel for agent payment programs (simulator) — every charge is matched to a mandate record within seconds; unmatched charges generate an auto-assembled dispute pack and an issuer-freeze request before the statement closes. PASS-CONDITIONAL(simulator) — fixture charge stream + fixture mandate ledger only, value_class "test"; nothing is filed with or requested of any issuer. Sole condition: full coordinated rollback = fenced condition — external anchor required before live/multi-party use.

charge_ingest · charge_status · program_summary · pack_get · freeze_request_get · operator_counter_sign · period_close · queue_read · pack_verify · review_mark

Test Bench: PASS-CONDITIONAL(simulator) — PASS-CONDITIONAL(simulator) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave3/bench-phantom.md) + delta re-bench (argus101-program-20261027/wave3/rebench-phantom.md, D1 CLOSED): builder suite 35/35 re-run clean by bencher; decisive fixture F-P13 re-derived from the bencher's own corpus (24 charges, 21 matched, 3 phantom classes, freeze assembled before period close). PASS-CONDITIONAL is not CERTIFIED — certified is false. Scope is exactly: local stdio, fixture charge stream + fixture mandate ledger, value_class 'test'; no real charge is reconciled and nothing is filed with or requested of any issuer. Sole condition (verbatim): full coordinated rollback = fenced condition — external anchor required before live/multi-party use. Advisories F-1/F-2 (non-blocking, spec-conformant): existence precedes verifiability in the fixed §4.3 order, and no MCP verb exists for MATCHED-receipt verify (spec §5 as written). Signing is an HMAC fixture stand-in for Herald RS256 (disclosed).

Usage: not collected — by design

Cyan Eclipse Capline

Sealed-policy cap bound proofs over fixture policies — a range proof over a sealed internal policy: the agent proves 'this spend is within my cap' to a merchant without revealing the cap or the policy. PASS-CONDITIONAL(local) — fixture policies/spends + fixture unaudited group parameters only, value_class "test"; no live merchant, no real policy proven. verdict-oracle disclosure: each verdict reveals one bit of ordering information about the sealed cap (whether the stated spend is above or below it); repeated proofs against one commitment narrow the cap by binary search (spec 2.3(c)). stated-spend disclosure: the stated spend is public to the verifier (it is the merchant's own price) (spec 11.4).

capline_policy_seal · capline_proof_build · capline_proof_status · capline_seal_revoke · capline_proof_present · capline_certificate_verify · capline_verifier_counter_sign · capline_certificate_get · capline_seal_list · capline_dispute_pack_get · capline_anomaly_queue_read

Test Bench: PASS-CONDITIONAL(local) — PASS-CONDITIONAL(local) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave3/bench-capline.md) + delta re-bench (argus101-program-20261027/wave3/rebench-capline.md, 45/45 checks PASS: D1 CLOSED, D2 CLOSED, D3 CLOSED, D4 accepted as documented deviation (not a defect), C4 DISCHARGED): builder suite 31/31 re-run clean by bencher at re-bench (27 original + 4 D1/D2/D3 regressions); decisive fixture F-C14 independently re-derived at bench and re-bench (see capabilities). PASS-CONDITIONAL is not CERTIFIED — certified is false. Scope is exactly: local stdio, fixture policies/spends, fixture unaudited group parameters, no live merchant, no real policy proven, value_class 'test'. Conditions (verbatim-listable): C1 §10 two-key gate — counsel clearance via FinTech Law packet AND independent crypto audit, both required, no dormant live path: live phases (real enterprise policy/cap, merchant-facing reliance, production parameters/audited cryptosystem, privacy/compliance marketing claims) require BOTH (i) counsel clearance through the existing FinTech Law packet AND (ii) an independent cryptographic audit of the §2.2 construction and this implementation; neither substitutes for the other. C2 cross-artifact cross-commitment fenced — external anchor required before live/multi-party use: full coordinated cross-artifact rollback requires an external anchor before live/multi-party use; v0.1 state is single-process local, anchored only by its signed store head + hash chains (re-bench §1 demonstrated a whole-store rollback to an earlier self-consistent snapshot is accepted locally — exactly this fenced condition). C3 deterministic fixture-seed prover randomness must be replaced/audited before any live phase: deterministic fixture-seed-derived prover randomness must be replaced by a CSPRNG (or an audited deterministic derivation under production key custody) before any live phase — seed compromise recomputes every nonce and blinding factor, including the cap. C5 "Capline" trademark screen still pending: trademark screen on the name "Capline" is still pending; the name remains a placeholder. Both §11.4 disclosures, verbatim: verdict-oracle disclosure: each verdict reveals one bit of ordering information about the sealed cap (whether the stated spend is above or below it); repeated proofs against one commitment narrow the cap by binary search (spec 2.3(c)). stated-spend disclosure: the stated spend is public to the verifier (it is the merchant's own price) (spec 11.4). Novel mechanism (SPEC §2.1, verbatim): a range proof over a sealed internal policy — the agent proves 'this spend is within my cap' to a merchant without revealing the cap or the policy. Signing is an HMAC fixture stand-in for Herald RS256 (HMAC Herald stand-in, disclosed on every artifact face as signature_scheme); a live phase would need real Herald signing and a re-bench.

Usage: not collected — by design

Cyan Eclipse Ward

Guardian veto verdicts + receipts over fixture actions; local/simulator only. PASS-CONDITIONAL (local/simulator) — a designated fixture guardian holds veto rights over sealed action classes, and every veto (or veto-window lapse) is receipted into the agent's own hash-chained record, recomputable offline from the exported pack. Ward issues verdicts and receipts; it does not enforce them. Rollback fence: the pure local file store caps any verdict at PASS-CONDITIONAL — an external anchor is required before any live or multi-party reliance. Counsel gate (SPEC §10) named and unbuilt: live runtime reliance and any legally recognised ward/guardian use require counsel clearance first.

ward_enrol · action_propose · action_withdraw · guardian_change · ward_status · ward_pack_get · clock_advance · ward_close · veto_cast · action_approve · guardian_accept_designation · verdict_get · pack_verify

Test Bench: PASS-CONDITIONAL (local/simulator) — Cyan Eclipse Ward — PASS-CONDITIONAL (local/simulator) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave4/bench-ward.md) PASS-CONDITIONAL with the guardian-change subsystem carved out under condition C1, + delta re-bench of v0.1.1 (argus101-program-20261027/wave4/rebench-ward.md): defects D1–D10 all CLOSED, guardian-change carve-out LIFTED, bencher delta harness 57/57, builder suite 37/37 re-run by the bencher at commit 35b00eb. PASS-CONDITIONAL (local/simulator) is not CERTIFIED — certified is false, and nothing beyond that verdict at its exact scope is claimed. Scope is exactly: local stdio, fixture action stream + fixture guardian only, value_class "test"; no real agent action is proposed, vetoed, executed, or stopped, and no production runtime is connected. Ward issues verdicts and receipts; it does not enforce them — enforcement is the consuming runtime's mandate layer. Rollback fence: the v0.1 store is a pure local file store, so any verdict is capped at PASS-CONDITIONAL — an external anchor is required before any live or multi-party reliance. Counsel gate (SPEC §10) named and unbuilt: live runtime reliance and any legally recognised ward/guardian use require counsel clearance first; no dormant live path exists.

Usage: not collected — by design

Cyan Eclipse Epitaph

Agent end-of-life cascade expiry over fixture identities; local/simulator only. PASS-CONDITIONAL (local/simulator) — a signed death/compromise record triggers cascade expiry: every outstanding mandate, alias, and delegation issued to the dead identity expires atomically and provably. Rollback fence: full coordinated rollback is a fenced condition — external anchor required before live/multi-party use. Counsel gate (SPEC §10) named and unbuilt for all live phases.

epitaph_check · cascade_receipt_get · cascade_verify · sweep_verify · death_record_present · dispute_file · dispute_adjudicate · identity_status · anomaly_queue_read

Test Bench: PASS-CONDITIONAL (local/simulator) — Cyan Eclipse Epitaph — PASS-CONDITIONAL (local/simulator) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave4/bench-epitaph.md) PASS-CONDITIONAL with conditions C1–C4, + delta re-bench of v0.1.1 (argus101-program-20261027/wave4/rebench-epitaph.md): defects D-EPI-1..4 all CLOSED, conditions C1–C4 satisfied by the closures, builder suite 38/38 re-run by the bencher at commit 0c0b027, no new defect found. PASS-CONDITIONAL (local/simulator) is not CERTIFIED — certified is false, and nothing beyond that verdict at its exact scope is claimed. Scope is exactly: local stdio, fixture identities and fixture registries only; no real identity, mandate, alias, or delegation is expired, revoked, or contacted, and no external registry is written to. A cascade receipt proves these fixture instruments expired under this signed record — never that a real agent "is dead", that a real compromise occurred, or that any real-world instrument is unenforceable. Rollback fence: cross-artifact cross-commitment (full coordinated rollback) is a fenced condition — external anchor required before live/multi-party use; the pure local file store caps any verdict at PASS-CONDITIONAL. Counsel gate (SPEC §10) named and unbuilt for all live phases.

Usage: not collected — by design

Cyan Eclipse Ancestor

Sealed model-ancestry chains + buyer policy checks over fixture chains/policies; local only. PASS-CONDITIONAL (local) — the full fine-tune/distillation ancestry chain is sealed at registration, and handshakes enforce buyer ancestry policies automatically. Ancestry is creator-declared and sealed at registration; sealing proves the declaration has not changed since registration — it does not independently prove the declared lineage. Rollback fence: an external anchor is required before live/multi-party use. Counsel gate (SPEC §10) named and unbuilt.

ancestry_check · chain_get · verdict_verify · chain_seal_verify · chain_register_seal · chain_amend · registration_status · anomaly_queue_read

Test Bench: PASS-CONDITIONAL (local) — Cyan Eclipse Ancestor — PASS-CONDITIONAL (local) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave4/bench-ancestor.md), + delta re-bench of v0.1.1 (argus101-program-20261027/wave4/rebench-ancestor.md: D1 + D2 CLOSED, new defect D3), + spot re-bench of v0.1.2 (argus101-program-20261027/wave4/spot-ancestor-012.md: D3 CLOSED, 22/22 binding probes, builder suite 39/39 reproduced). PASS-CONDITIONAL (local) is not CERTIFIED — certified is false, and nothing beyond that verdict at its exact scope is claimed. Scope is exactly: local stdio, fixture chains + fixture policies only, value_class "test"; no real model, registry, or counterparty is contacted and no live handshake is mediated. Declaration label: ancestry is creator-declared and sealed at registration; sealing proves the declaration has not changed since registration — it does not independently prove the declared lineage. Gap-verdict condition: a verified gap verdict binds to the chain + seal + sealed head + gap kind only; it does NOT identify which link is missing or unverifiable, and must not be presented as evidence about the exact presented link set or as proof two presentations are the same copy. Rollback fence: the pure local file store caps any verdict at PASS-CONDITIONAL — an external anchor is required before live/multi-party use. Counsel gate (SPEC §10) named and unbuilt.

Usage: not collected — by design

Cyan Eclipse Greenwash

Emission-time claim binder for agent output over a fixture emission stream; local/fixture only. PASS-CONDITIONAL (local/fixture) — every sustainability claim an agent emits is bound to its source-data pointer at emission time; unbound claims are flagged in the message itself. A binding proves a pointer existed and resolved at emission time — NOT that the claim is true. Rollback fence: an external anchor is required before live/multi-party use. Counsel gate (SPEC §10) named and unbuilt.

emission_submit · claim_status · ledger_get · binding_verify · pack_get · annotation_add

Test Bench: PASS-CONDITIONAL (local/fixture) — Cyan Eclipse Greenwash — PASS-CONDITIONAL (local/fixture) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave4/bench-greenwash.md) PASS-CONDITIONAL with conditions C1–C4, + delta re-bench of v0.1.1 (argus101-program-20261027/wave4/rebench-greenwash.md): defects D-GW-1..4 all CLOSED, conditions C1–C4 satisfied (C2 via the labelling branch), builder suite 34/34 re-run by the bencher at commit ba07801, no new defect found. PASS-CONDITIONAL (local/fixture) is not CERTIFIED — certified is false, and nothing beyond that verdict at its exact scope is claimed. Scope is exactly: local stdio, fixture emission stream with structured claim objects only, value_class "test"; no real agent, channel, publication, or marketing surface is touched, and no free-text claim detection in arbitrary prose is claimed or implied. Pointer limit: a binding proves a source-data pointer existed and resolved at emission time — NOT that the claim is true, not that the source data supports the claim, and not that the source data is accurate, current, or complete. Rollback fence: pure local file store caps any verdict at PASS-CONDITIONAL (local/fixture) — external anchor required before live/multi-party use. Counsel gate (SPEC §10) named and unbuilt for any real-world claim publication. Trademark screen on the working name pending.

Usage: not collected — by design

Cyan Eclipse Agesure

Band-only age attestation sealer/verifier over fixture attestations; local/fixture only. PASS-CONDITIONAL (local/fixture) — band-only attestation (18+ / 21+) that reveals nothing else — no birthdate, no document — sealed to the requesting service's mandate. It is not an identity claim. Rollback fence (C1): external anchor required before live/multi-party use. Counsel gate (C2, SPEC §10) named and unbuilt: any use involving real people.

proof_request · proof_verify · proof_status · revocation_present · issuance_pack_get · attestation_register

Test Bench: PASS-CONDITIONAL (local/fixture) — Cyan Eclipse Agesure — PASS-CONDITIONAL (local/fixture) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave4/bench-agesure.md) PASS-CONDITIONAL with conditions C1–C7, + delta re-bench of v0.1.1 (argus101-program-20261027/wave4/rebench-agesure.md: D1–D8 all CLOSED, new defect D9 + condition C8), + spot re-bench of v0.1.2 (argus101-program-20261027/wave4/spot-agesure-012.md: D9 CLOSED, C8 DISCHARGED, builder suite 35/35 reproduced). PASS-CONDITIONAL (local/fixture) is not CERTIFIED — certified is false, and nothing beyond that verdict at its exact scope is claimed. Scope is exactly: local stdio, fixture attestations only, value_class "test"; no real person is checked and no real age-verification capability is claimed. A proof states that an issuer attested a band for a pseudonymous subject ref, sealed to one mandate; it is not an identity claim and does not prove the issuer's underlying band determination was correct. Standing conditions: C1 rollback fence (pure local file store caps the verdict at PASS-CONDITIONAL — external anchor required before live/multi-party use), C2 §10 counsel gate (named, unbuilt — any use involving real people), C5 issuer key separation, C7 name screen still bind.

Usage: not collected — by design

Cyan Eclipse Fairlend

Decision-evidence sealer + audit recompute engine over fixture decisions; local/fixture only. PASS-CONDITIONAL (local/fixture) — each decision seals its feature set, model version, and adverse-action reason codes at decision time, making later disparate-impact audits a recompute. Fairlend never decides. D-11 NAMED: §7.3 identifier fence evadable by token-split spellings — must close before any non-fixture scope.

decision_seal · model_register_fixture · reason_table_register_fixture · decision_get · decision_retract · seal_verify · audit_define · audit_run · audit_receipt_get · audit_pack_get · decision_export

Test Bench: PASS-CONDITIONAL (local/fixture) — Cyan Eclipse Fairlend — PASS-CONDITIONAL (local/fixture) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave4/bench-fairlend.md) + delta re-bench of v0.1.1 (argus101-program-20261027/wave4/rebench-fairlend.md: D-1..D-6, D-8, D-9, D-10 CLOSED; D-7 closed for its probed class with residual Low defect D-11 NAMED; builder suite 36/36 reproduced). PASS-CONDITIONAL (local/fixture) is not CERTIFIED — certified is false, and nothing beyond that verdict at its exact scope is claimed. Scope is exactly: local stdio, fixture decisions only, synthetic cohorts; no real applicant, no real credit data, no real model, and no lender system is touched, and Fairlend makes no credit decision and influences none. A Fairlend seal proves what was recorded at decision time — the presented feature set, model version, and reason codes, counter-signed then; it does not prove the decision was fair, lawful, unbiased, or correct. Standing condition D-11 (NAMED): the SPEC §7.3 identifier fence is still evadable by token-split spellings (tax_id, ein, gov_id, surname, home_street etc. seal today when they should refuse); D-11 must close before any non-fixture scope. Rollback fence and the SPEC §10 fair-lending counsel gate (named, unbuilt) also bind.

Usage: not collected — by design

Cyan Eclipse Awardline

Grant-term spending evidence over fixture awards; local/simulator only. PASS-CONDITIONAL (local/simulator) — award terms sealed once; each spend sealed against a term as it happens, hash-chained per award; reports recompute from those receipts. An Awardline report is a recomputation from sealed receipts, not an audit, not an allowability determination, and not agency acceptance. Counsel/govcon gate (C1, SPEC §10) named and unbuilt: any real-award reliance.

awardline_terms_register · awardline_spend_seal · awardline_status · awardline_report · awardline_period_close · awardline_export · awardline_verify

Test Bench: PASS-CONDITIONAL (local/simulator) — Cyan Eclipse Awardline — PASS-CONDITIONAL (local/simulator) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave4/bench-awardline.md) + delta re-bench of v0.1.1 (argus101-program-20261027/wave4/rebench-awardline.md: D1–D5 all CLOSED, condition C4 SATISFIED, builder suite 26/26 reproduced). PASS-CONDITIONAL (local/simulator) is not CERTIFIED — certified is false, and nothing beyond that verdict at its exact scope is claimed. Scope is exactly: local stdio, fixture award terms + fixture spend records only, value_class "test"; no real award, real grantee, real funds, real accounting system, or government system is touched. Load-bearing label: an Awardline report is a recomputation from sealed receipts, not an audit, not an allowability determination, and not agency acceptance. Standing conditions: C1 §10 counsel/govcon gate (unbuilt, binds any real-award reliance), C2 external anchor (caps the verdict at PASS-CONDITIONAL), C3 real Herald RS256 (required with re-bench before any live phase), C5 name screen still bind.

Usage: not collected — by design

Cyan Eclipse Insurepack

Continuous cyber-insurance evidence assembly over fixture records; local/simulator only. PASS-CONDITIONAL (local/simulator) — a live underwriter-facing evidence API fed by bench verdicts, incident records, and control receipts — premium-relevant changes emit change receipts to the insured. Evidence assembly only; no underwriting. Amended Condition 1′: a single-pack VERIFY_PASS is not proof of completeness of earlier transitions. Counsel gate (SPEC §10) named and unbuilt: any live phase.

insurepack_profile_register · insurepack_input_present · insurepack_assemble · insurepack_pack_get · insurepack_verify · insurepack_changes_list · insurepack_export

Test Bench: PASS-CONDITIONAL (local/simulator) — Cyan Eclipse Insurepack — PASS-CONDITIONAL (local/simulator) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave4/bench-insurepack.md) + delta re-bench of v0.1.1 (argus101-program-20261027/wave4/rebench-insurepack.md: DEFECT-1..3 all CLOSED, Condition 1 amended to Condition 1′ by the bench, builder suite 28/28 reproduced). PASS-CONDITIONAL (local/simulator) is not CERTIFIED — certified is false, and nothing beyond that verdict at its exact scope is claimed. Scope is exactly: local stdio, evidence assembly over fixture verdict/incident/control records only, value_class "test"; no live API is built, no underwriter is served, no insurer system is touched, and no real insured, real incident, or real policy enters a pack. Load-bearing disclosure (SPEC §11.1): the evidence pack is insured-side disclosure tooling; underwriting reliance is the insurer's process, not Insurepack's claim. Amended Condition 1′ binds verbatim: a single-pack VERIFY_PASS attests the pack, its own transition, and the internal integrity of the presented stream; it must NOT be relied on as proof of completeness of transitions earlier than the pack's immediate predecessor (a fully re-signed, re-linked rewritten stream is indistinguishable there); full-history reliance requires per-pack chain verification (which the export transcript now performs) or an external anchor. Rollback fence and the SPEC §10 counsel gate (named, unbuilt) also bind.

Usage: not collected — by design

Cyan Eclipse Ballotbox

Ballot evidence system for agent electorates over fixture electorates; local/simulator only. PASS-CONDITIONAL (local/simulator) — ballots sealed to mandate-eligibility proofs: only agents holding a valid governance mandate at vote time are counted, and the tally recomputes publicly. Recompute proves tally-vs-sealed-ballots, not electorate legitimacy. Counsel gate (C4, SPEC §10) named and unbuilt: any real binding election.

election_create · electorate_seal · election_open · ballot_cast · ballot_status · election_close · election_void · election_pack_get · tally_compute · tally_verify · tally_receipt_get · tally_receipt_counter_sign

Test Bench: PASS-CONDITIONAL (local/simulator) — Cyan Eclipse Ballotbox — PASS-CONDITIONAL (local/simulator) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave4/bench-ballotbox.md) + delta re-bench of v0.1.1 (argus101-program-20261027/wave4/rebench-ballotbox.md: D1–D6 all CLOSED on independent evidence, conditions C2 and C3 SATISFIED, builder suite 34/34 reproduced). PASS-CONDITIONAL (local/simulator) is not CERTIFIED — certified is false, and nothing beyond that verdict at its exact scope is claimed. Scope is exactly: local stdio, fixture electorate + fixture governance mandates only, value_class "test"; no real person, company, union, public body, or live agent fleet votes in v0.1.1, no result binds anyone, and no real decision is taken here. The legitimacy limit binds verbatim: recompute proves tally-vs-sealed-ballots — that the published totals are exactly the function of the sealed ballots and their vote-time eligibility proofs; it does not prove electorate legitimacy (who chose the electorate, whether the mandate issuer was itself authorised, whether voters were coerced, or whether the question was fair), and Ballotbox claims none of them. Standing conditions: C1 (HMAC cap — no public/RS256/JWS claim until real Herald signing), C4 (§10 counsel gate, named and unbuilt — any real binding election), C5 (metering-only pricing), C6 (rollback fence) still bind.

Usage: not collected — by design

Cyan Eclipse Exportcheck

Screening-evidence sealer over fixture lists only; local/fixture only. PASS-CONDITIONAL (local/fixture) — every verdict seals the exact list vintage it was computed against (snapshot id, vintage label, content hash) plus the sealed rule-set version, so a clearance remains provable after the lists change. A fixture CLEAR authorises no shipment, transaction, or filing. Export-control counsel gate (SPEC §10) named and unbuilt: any real list, real party, or reliance.

snapshot_load_fixture · snapshot_list · ruleset_register_fixture · party_screen · verdict_rescreen · verdict_get · verdict_verify · vintage_diff · verdict_pack_get · verdict_export

Test Bench: PASS-CONDITIONAL (local/fixture) — Cyan Eclipse Exportcheck — PASS-CONDITIONAL (local/fixture) — Verity bench of v0.1.0 (2026-10-27, argus101-program-20261027/wave4/bench-exportcheck.md: PASS-CONDITIONAL (local/fixture), defects D1/D2, conditions C1-C3) + collision-screen fix rounds and spot re-benches rounds 1-6 closing at wave4/spot-exportcheck-022.md (v0.1.7 @5b0ab96: D3 CLOSED, survivors-in-claim 0 forms / 0 glyphs over an exhaustive 2,749-glyph / 9,836-form sweep, determinism 55/55 processes one digest, invariance byte-identical vs v0.1.2, C4 DISCHARGED) — final product verdict PASS-CONDITIONAL (local/fixture) (spot-022 verdict: PASS, scoped to the D3/C4 condition only; the C3 rollback fence is undischarged and caps the product verdict at PASS-CONDITIONAL (local/fixture) until an external anchor exists). PASS-CONDITIONAL (local/fixture) is not CERTIFIED — certified is false, and nothing beyond that verdict at its exact scope is claimed. Scope is exactly: local stdio, fixture parties against invented fixture snapshots only, value_class "test"; no real list is loaded, fetched, or reproduced, no real party is screened, and no verdict is relied on for anything. Collision screening (local/fixture scope): fixture sets loaded into Cyan Eclipse Exportcheck are collision-screened against symbol-, format- and homoglyph-decorated forms of listed names under SPEC Q-E3 as amended at v0.1.7 — assignment, identity and transmitted readings as there defined (bench: spot-exportcheck-022, survivors-in-claim 0). Standing disclosure: fixture sets are invented test data; a fixture CLEAR states only that this invented party record did not match this invented snapshot under this sealed rule-set version, as recomputed by spec 6.3 — it authorises no shipment, transaction, or filing. No live claim is made; no certification is claimed beyond the stated local/fixture bench scope. Standing conditions: C2 (from bench deviation 2): party_screen never links (prior_verdict_id: null) — a repeat party_screen under a newer snapshot produces an unlinked but immutable, vintage-sealed verdict; chain claims rest on verdict_rescreen + pack chain receipts only, and no listing or doc may imply party_screen chains. C3 (rollback fence, builder-stated, bencher-confirmed): verdict caps at PASS-CONDITIONAL(local/fixture) until an external anchor exists; pure local file store + signed head/SET anchors only. No test-bench scorecard was issued for this verdict and none is claimed (card_url null).

Usage: not collected — by design

Cyan Eclipse Test Bench

The certification harness behind every verdict on this shelf: adversarial suites run over the wire against bench-owned fixtures — homoglyph typosquats, lying backends, duplicate-event batches, schema fuzzing — with FAILED cards retained append-only, never overwritten. It runs in-house against the whole fleet today and is productizing as a CLI + hosted run service (slate T1). The paid tier — a Verity-run certification carrying the seal — is an owner decision, not yet offered.

In-house artifact — productization in progress; no public Test Bench card yet, and none claimed.

Usage: not collected — by design

Cyan Eclipse Stocktake

Catalog / manifest / storefront consistency auditor: one scorecard ledger checked against every surface that repeats a verdict, with machine-safe fields synced and prose verified, never rewritten. It is the drift checker that has closed every Eclipse publish pass at drift=0 — running in-house today as sync_verdicts.py, productizing as a CLI + CI action (slate T4).

In-house artifact — productization in progress; no public Test Bench card yet, and none claimed.

Usage: not collected — by design

Cyan Eclipse Press

Agent-native static-site publish pipeline: deploy zips with normalized permissions and real file mtimes — the GoDaddy 403 and PHP-opcache lessons, encoded — published to a verified shelf manifest. It publishes thewhitereport.com and this Bazaar in production today (deploy-pipeline/publish.py) and is productizing as a standalone pipeline (slate T3).

In-house artifact — productization in progress; no public Test Bench card yet, and none claimed.

Usage: not collected — by design

Herald Card Kit

Agent Card issue/verify toolkit for A2A builders: the Herald JWS card tooling (cards.py, jwscore.py) that already runs inside the Cyan Eclipse Herald MCP, productizing as a CLI + library (slate T7). Issue signed Agent Cards and verify a counterparty's card offline, with receipts — no call home.

In-house artifact — productization in progress; no public Test Bench card yet, and none claimed.

Usage: not collected — by design

Cyan Eclipse Trust MCP

One call answers whether a bench result is real, who issued it, at what scope — and whether it is anchored on-chain, where.

trust_status · trust_verify_receipt · trust_get_bench_verdict · trust_get_validation_record · trust_resolve_agent

Test Bench: PASS(local) — PASS(local) — Verity independent bench of v0.1.0 (2026-10-27, trust-discovery-20261027/BENCH-REPORT.md): first bench PASS-CONDITIONAL(local) with all fail-gates PASS (real Anchor batch b18e97a2 end-to-end, forged-pointer fixture, verdict fidelity, malformed battery); condition F1 (validation verify_recipe canonicalization wording) + F2/F3 closed by delta bench — final verdict PASS. Scope: stdio + HTTP twin over the bundled join index; index recomputed from sources (Bazaar verdict feed, Herald ledger, Anchor batches/proofs, ERC-8004 publisher ledger) by a fail-closed builder. Hosted PHP twin: static parity PASS at bench; execution verified by live smoke at deploy 2026-10-27.

Usage: not collected — by design

Cyan Eclipse Ledger MCP

Betting records and responsible-gambling controls: limits that can fall instantly but never rise without a wait and a fresh confirmation, absolute self-exclusion, honest cash P&L that never counts a bonus bet as money. Not a sportsbook. No bets are placed here.

limit_set · limit_confirm · exclusion_start · jurisdiction_check · bet_record_add · pnl_summary · reality_snapshot · odds_compare · slip_prepare · resources_get

Test Bench: CERTIFIED(local) — 21/22 checks · fixtures only · not a sportsbook

Usage: not collected — by design