Use when curating, refreshing, expiring, calibrating, or equivalence-testing evidence and publication inputs for the active Safety Score model. Not for unrelated scoring models or copy-only changes.
The English summary is being prepared. Showing the original description.
See how to installReview the instructions the agent will receive before you install.
Use for active-model curation and score-affecting refreshes, not other scoring models or copy-only edits. Stable safety-score-v9 paths/commands do not identify the current methodology version; read shared/lib/methodology-versions/current-version.json.
Select evidence curation, weekly expiry sweep, mechanism/shock/protocol producer refresh, DDRR calibration, equivalence testing, or methodology change before loading branch context. Never infer evidence from a schema gap.
shared/data/safety-score-v9/AGENTS.md; Active model, Canonical Publication, and equivalence gate selection.shared/data/safety-score-v9/methodology-policy-candidate-v1.json for policy and scripts/lib/automation-registry.mjs for artifact ownership.assessV9Publication's partial-publication allowance in worker/src/lib/safety-score-v9/publication-assessment.ts.shared/data/safety-score-v9/evaluation-build-manifest-v1.ts.safety-score-v9:missing-data-registry and safety-score-v9:curation-worklist with the linked capture/output arguments, claim-group decisions, current primary evidence, and named promote/reject/defer triggers. safety-score-v9:expiry-queue is preventive, not a complete missing-evidence inventory. For the weekly sweep, also read weekly close, run safety-score-v9:live-withheld -- --replay <same-replay> --output <path>, and drain DEP/RESV, pre-expiry and counterfactual lanes; the other queues omit live-to-fallback grade drops.method-unsupported; exact UTC timestamps enter at their instant (isMechanismOverlayCurrent in worker/src/lib/safety-score-v9/extension-mechanism.ts). Both retain original-start expiry; component-only reviews never renew siblings or an expired whole overlay.npm run research:dwellir-rpc -- under pinned-evidence rules. Cite the keyless URL, block and timestamp provenance, never latest.--rederive-current-redemption is only for expiry/current-curation scenarios, never accepted/embedded-registry captures or --registry-ref; registry mismatch is only for attributed code/curation comparison. Require two independent captures at least 90 minutes apart, normalized score-output equality for neutral changes or prospective reviewed grade/status movers. Malformed/duplicate cards/manifests fail. Manually close absent declarations/census/same-grade findings; unexpected movers stop release. Independently verify activation identity and source-revision/registry/clock/accepted-enrichment reproduction: an empty normalized diff is not bit-identical identity proof.The assigned verifier runs applicable checks after writers finish; report unrun requirements rather than claiming proof.
npm run safety-score-v9:replay, npm run safety-score-v9:diff, npm run safety-score-v9:movers, with branch-owned capture/output arguments.npm run generate:safety-score-v9-evaluation-build, then npm run check:generated-artifacts -- --only=safety-score-v9-evaluation-build.npx vitest run worker/scripts/__tests__/replay-safety-score-v9.test.ts worker/scripts/__tests__/diff-safety-score-v9-replays.test.ts shared/types/__tests__/safety-score-v9-overlays.test.ts.Report work/assets, sourceGeneration, baseInputGenerationId, clockSec, evidence, replay/diff/movers, artifact ID, methodology targets when applicable, actual verifier results, post-deploy acceptance or pending evidence, and risks.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Review the retained annotation candidate queue and curate per-coin editorial history. Use weekly during active event periods, monthly otherwise, or on queue-health warnings; this does not restore the retired live chart overlay.
The English summary is being prepared. Showing the original description.
Generate a weekly changelog entry from git history, filtering operational noise and producing editorial summary, field notes, statistics, and a bounded commit manifest. Stops for review unless commit was explicitly requested.
The English summary is being prepared. Showing the original description.
Research and populate tracked stablecoin compliance sidecars for the U.S. GENIUS Act, EU MiCA, or both. Use when adding or auditing authorization, applicability, pathway, or disclosure evidence.
The English summary is being prepared. Showing the original description.
Use when adding, reviewing, squashing, deploying, or rolling back Pharos D1 migrations or schema cleanup.
The English summary is being prepared. Showing the original description.
Drain the Pharos Dependency Map coverage-audit queues into evidence-refreshed edges, reviewed target dispositions, or documented deferrals. Use monthly after weekly production dependency audits, after dependency coverage drops, or when reviewing adapter gaps, material unlinked slices, symbol leads, duplicate/split groups, or unverifiable published edges.
The English summary is being prepared. Showing the original description.
Update Pharos funding donations by reconciling inbound transfers, rejecting wallet self-activity and spam, pricing receipt-time value, and appending user-approved rows.
The English summary is being prepared. Showing the original description.