菜单
Fixed-topology, discovery-only extension for Gaia curation. Atomically checkpoints each candidate transition, retries only the failed field, and stops at human L4 review.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Helper for release operations, compilation of changelogs between tags, version lockstep syncing, and promotion of GitHub releases from canary/pre-release to latest/production. Use when the user asks to "compile the changelog", "promote release to latest", "tag a release", or do version release operations.
Push committed changes and create a draft PR on GitHub. Use when you have staged changes ready to push and want to open a draft PR for early feedback or documentation.
Consult this skill BEFORE proposing a fix to any CLI gap or schema gap that touches evidence, trust magnitude, evidence types, evidence grades, the TM formula, or star-level promotion gates. Also read it whenever an issue mentions citations, stars, commits, contributors, views, reviewers, percentile, arxiv, github-stars-own, repo-own, social-signal, benchmark-result, peer-review, self-attestation, fusion-recipe, verifier-attestation, or proxy-containment — those are the 10 canonical evidence types and the raw inputs that feed their formulas. Also read it when asked whether a skill qualifies for a given star level (0★–6★): the Star Bar and promotion gates live in `META.md`, not in the TM methodology page. Reading both prevents proposing a calibration that satisfies the TM grade threshold but violates a Star Bar installability or link requirement.
Phase 4 of the Gaia evidence verification pipeline. Checks every URL in the evidence data lake for HTTP liveness — catches dead links (404, 403, timeout, connection errors) before evidence is ingested into the registry. Use this skill when: running the full /ev-pipeline, asked to "validate links", "check evidence URLs", "run link validation", "find dead links in the data lake", "check HTTP status of evidence sources", or "verify evidence links still work". Also useful standalone when you suspect evidence sources have gone stale after a long gap between pipeline runs. Wraps validate_sources.py, which uses Firecrawl CLI to scrape each unique URL and records pass/fail status in a markdown report. Requires Firecrawl CLI to be installed and authenticated.
Top-level orchestrator for the Gaia Registry evidence verification pipeline. Run this when you want to do a full evidence pass — collecting raw sources, verifying live GitHub star counts, adversarially auditing for quality/formatting issues, and checking that every URL is reachable — all in one coordinated sequence. Trigger phrases: "run the evidence pipeline", "full evidence pass", "verify evidence", "evidence verification pipeline", "run ev-pipeline", "check all evidence", "audit the data lake", "evidence quality sweep", "refresh evidence", "validate evidence sources". Also aliased as /evidence-verification-pipeline. Use the individual sub-skills (ev-collection, ev-star-verification, ev-adversarial-audit, ev-link-validation) only when you need to re-run one phase in isolation; for end-to-end work, always start here.
Canonical entry point for the full Gaia evidence verification pipeline. Invoke as /evidence-verification-pipeline or /ev-pipeline — both names trigger this skill. Use whenever you need to run all four evidence phases end-to-end: collecting raw evidence from the data lake, verifying live GitHub star counts, running adversarial auditing for noise and URL errors, and checking link health via Firecrawl. Also use when someone says "run the evidence pipeline", "verify the evidence lake", "audit evidence", "run ev-pipeline", "full evidence check", "pipeline run", "prepare evidence for ingestion", or "refresh the data lake". This is the pre-ingestion step — run it before importing any evidence into the registry.
| name | gaia-curate-chain |
| description | Fixed-topology, discovery-only extension for Gaia curation. Atomically checkpoints each candidate transition, retries only the failed field, and stops at human L4 review. |
| version | 3.0.0 |
| argument-hint | <source-or-small-batch> |
Read ../gaia-curate/CURATION-CORE.md; it owns the lifecycle and packet. Use this extension when recoverability matters more than throughput.
Persist a run ledger under generated-output/curate-discovery/<run-id>/run.json containing the core contract digest, generic snapshot digest, source cursor, active candidate ID, completed candidate IDs, deferred rows, and next instruction. Write to a temporary file, validate it, then atomically rename it; never leave a partially written checkpoint.
Process exactly one candidate at a time. Validate its discovery-packet-v1 after every transition. Advance only when the current transition validates and its input digest matches the checkpoint. On resume, revalidate contract and generic snapshot digests; preserve raw source records but move stale mappings to DEFER.
Route only the failing candidate and field back to its worker. Allow at most two retries for transient fetch failure and one Luna High/equivalent repair for invalid structured output. A deterministic schema, hash, or generic-reference failure is not blindly retried.
If retries are exhausted or output remains invalid/ambiguous, emit DEFER with the stable core reason code and exact resume instruction: candidate ID, failed field, validator code, source cursor, and required operator action. Never synthesize or infer a decision.
The chain ends at L4 human review. It does not collect evidence, calculate trust, mutate the registry, regenerate docs, create intake batches, commit, push, or open a PR.
After an L4 decision, write a handoff record; do not silently end the caller:
skills.yml intake schema, then require the operator to run gaia push --from-file skills.yml or use the issue form./gaia-draft-curate: preserve batchId and issue/PR links, return them as needs-evidence to /ev-pipeline, then hand verified rows to a maintainer on review/meta/* for the CLI-only meta-shift in CONTRIBUTING.md §1C. Do not submit the same intake twice.Record the route, artifact path, remaining gate, and exact next command. This compatibility handoff is data only; the chain remains discovery-only.