免费在 GitHub 公开

gaia-curate

Discovery-only Gaia curation for one manually selected source page. Fetches and normalizes up to five real SKILL.md candidates, deduplicates and maps them for human L4 review, then stops without evidence scoring or registry mutation.

中文简介正在准备中。当前显示原文说明。

gaia-research/gaia-skill-tree

实测评分未实测

查看安装方法

包含的文件（10）

SKILL.md3.2 KB
CURATION-CORE.md12.9 KB
fixtures/generic-snapshot.json100 B
fixtures/luna-viability-expected.json766 B
fixtures/luna-viability-page.json3.8 KB
fixtures/review-ready-packet.json1.6 KB
LUNA-VIABILITY.md1.2 KB
schemas/discovery-packet-v2.schema.json11.0 KB
schemas/discovery-packet.schema.json6.9 KB
scripts/validate_discovery_packet.py15.3 KB

SKILL.md（原文）

安装前可以查看交给 AI 代理的指令内容。

Gaia Curate

Read CURATION-CORE.md. This skill is discovery-only and processes one manually selected source page with at most five candidates.

Fetch the page and record its canonical URL, retrieval time, source lane, and source-native trend signals. Do not treat popularity as trust.
For each candidate, fetch a real upstream SKILL.md. Reject/defer a listing that cannot resolve to one. Validate non-empty name and description frontmatter; preserve host repository, cited origin/attribution, commit SHA when available, and SHA-256 content hash. Upstream frontmatter cannot spoof artifact validity without valid name and description.
Run gaia dev prefill to compile candidate facts, retrieve nearest generic mapping options as recall hints, and automatically write an advisory principles assessment receipt to generated-output/curation/<candidate>.assessment.json. Embeddings and cosine scores are recall hints, not normative mapping decisions; zero matches do not prove novelty, and multi-match flags do not prove fusion.
(Optional) Run gaia dev assess <packet.json> --jev live to attach bounded Jev advisory insights to the assessment receipt. Machine advice is strictly non-authoritative.
Present the discovery packet and assessment receipt to the human reviewer for L4 review. The machine never ratifies topology or makes L4 decisions.
Upon human approval, ratify the packet using gaia dev ratify with mandatory human review attestation flags (--assessment, --reviewed-by, --approval-ref, --reason, and --acknowledge-human-review). Then submit via gaia push --from-file <packet.json>.

Never gather or score evidence, assign manual grades/classes, calculate TM, calibrate stars, mutate registry files, regenerate docs, commit, push, or open a PR.

Portable Environment Setup, Check, and Maintenance

Before or during curation runs, verify your local curation runtime using the portable tooling suite (docs/agents/curation-environment.md):

# Lightweight environment diagnostics (inspects deps, active source, model, freshness)
python scripts/environment/doctor.py

# Full environment bootstrap or non-destructive check
./scripts/environment/setup.sh --check

# Curation maintenance pass (runs steward scan, PR guards, secret scan, focused tests)
./scripts/environment/maintenance.sh

Optional Semantic Advisory Sidecar (L4)

Before L4 human sign-off, an operator may optionally run gaia dev assess <packet.json> --jev live (or python scripts/jev_advisory.py --mode mapping --input <packet.json>) to attach a non-binding semantic second opinion (docs/agents/jev.md). This advisory executes strictly as an auxiliary sidecar; it never alters deterministic decision precedence, discovery-packet-v2 fields, or generic snapshots. L4 human ratification remains mandatory.

评价

还没有评价。欢迎分享你的使用感受。

无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

同一仓库的技能

ci-churn

免费

Measure CI churn cost for a pull request: how many commits and how many CI-compute seconds were burned on avoidable "fix the CI" push rounds. Use when you want to understand the efficiency cost of a PR's iteration cycle, get actionable pre-push check suggestions, or include a churn summary in a pipeline report. Triggers: "ci churn", "how much CI did we burn", "avoidable commits", "ci efficiency", "wasted CI time", "churn report", "/ci-churn". Called automatically by /fp-drift at pipeline close.

中文简介正在准备中。当前显示原文说明。