Skip to content

gaia-research/gaia-skill-tree / .agents/skills/gaia-ingest

gaia-ingest

Verified

Atomically ingest one already-verified evidence row into one named Gaia skill using the CLI, then appraise Trust Magnitude and propose—never silently apply—a calibration. Use after /ev-pipeline has verified an approved intake source, or whenever a curator says "ingest evidence", "add verified evidence", "gaia-ingest", or "turn this source into registry evidence".

FileKindSize
SKILL.mdskill md5.2 KB
---
name: gaia-ingest
description: >-
  Atomically ingest one already-verified evidence row into one named Gaia skill
  using the CLI, then appraise Trust Magnitude and propose—never silently
  apply—a calibration. Use after /ev-pipeline has verified an approved intake
  source, or whenever a curator says "ingest evidence", "add verified evidence",
  "gaia-ingest", or "turn this source into registry evidence".
version: 1.0.0
argument-hint: "<contributor/skill> <verified-source>"
---

# Gaia Ingest

Safely turn **one verified evidence source** into one CLI-authored evidence row.
This is the mutation bridge after evidence verification; it is not discovery,
source research, or a bulk importer.

## Required inputs

- Named skill ID: `contributor/skill-id`
- Source URL verified by `/ev-pipeline` link validation
- Evidence Type and only the numeric payload that type supports
- Attribution scope: `named`, `generic`, or `suite`
- Factual notes describing exactly what the source proves

Refuse the write when any input is absent, the source is dead, the evidence
belongs to a different capability, or the requested payload would be guessed.

## Preflight

1. Confirm the named skill exists and inspect its current rows:

   ```bash
   gaia dev list --named --json
   sed -n '1,120p' registry/named/<contributor>/<skill-id>.md
   ```

2. Confirm the source was verified by the current evidence run. For GitHub,
   also fetch live facts with `gh api`; for social sources, scrape the source
   and preserve its public view count; for benchmarks, require the complete
   reproducibility payload.

3. Reject duplicate `(Evidence Type, canonical source URL)` rows. Do not use
   a suite-wide source as full-strength component evidence. A component must
   have implementation-specific proof; suite-wide proof belongs on the suite
   capstone or is recorded as scoped supplemental context. Never add a synthetic
   `self-attestation` row solely to push a suite component past the 49.9 TM
   shared-repository baseline cap (see [`docs/agents/curation-authority-map.md`](../../../docs/agents/curation-authority-map.md)).

4. Enforce #2473 social vs proxy eligibility:
   - `social-signal` admits both eligible first-party (author/maintainer) and
     third-party demonstrations that substantively cover the target skill
     (`sameAsAuthor = 1.0`, `publishCooldownDays = 0` — no 30-day publication hold).
     Require `views >= 1000` (below 1,000 scores 0 TM), record `creatorRelation`
     (`first-party` or `third-party`) and `relevanceNote` in notes/metadata,
     reject unrelated uploads, and respect `(type, url)` dedup, the 3-row
     `1.0 / 0.5 / 0.25` plateau, and the 80 TM hard cap.
   - `proxy-containment` strictly requires a qualifying external, non-first-party
     repository (`--external-stars` >= 10,000) whose `owner/repo` is distinct from
     the skill's contributor, own/canonical repo, and `suiteRef` repository.

5. Choose only a schema-supported Evidence Type. Never use the obsolete
   `--class` flag. Use `gaia dev evidence --help` if the payload is unclear.

## CLI-only mutation

Run exactly one command for the one evidence row. Always use `--no-build`;
callers such as `/gaia-ingest-batch` perform the single final build.

```bash
GAIA_OPERATOR_OVERRIDE=1 gaia dev evidence <contributor/skill-id> "<source-url>" \
  --type <evidence-type> <supported-numeric-flags> \
  --notes "<factual scope and verification note>" \
  --source-started-at YYYY-MM-DD \
  --no-build
```

Payload examples:

```bash
# Eligible first-party or third-party video demonstrating the skill; views must be visible on the scraped source (no 30-day hold).
gaia dev evidence firecrawl/web-scrape-integration "https://www.youtube.com/watch?v=..." \
  --type social-signal --views 8510 \
  --notes "Third-party tutorial explicitly demonstrates Firecrawl page scraping." \
  --source-started-at 2025-07-20 --no-build

# Concrete upstream skill file; use a blob URL, live star count, and real
# number of skills sharing the repository.
gaia dev evidence firecrawl/web-scrape-integration "https://github.com/firecrawl/skills/blob/main/skills/firecrawl-build-scrape/SKILL.md" \
  --type github-stars-own --stars 42 --skill-count-in-repo 5 \
  --notes "Official upstream implementation source; star count verified via GitHub API." \
  --source-started-at 2026-04-08 --no-build
```

For `benchmark-result`, provide every reproducibility flag required by
`gaia dev evidence --help`; never convert a vendor benchmark claim into a
percentile without an independently verifiable benchmark source.

## Appraise, then stop

After the write:

```bash
PYTHONPATH=src python3 scripts/trust_appraise.py --skill <contributor/skill-id>
```

Report the new TM, Overall Trust Grade, per-row score, and whether the source
is implementation-specific. Propose a calibration only when Trust Magnitude clears
the gate and the Star Bar is met (the per-star Evidence Floor is retired under
Yggdrasil II — TM is the sole promotion gate). Do **not** run `gaia dev calibrate` without explicit
operator approval.

## Finalization

For a one-row standalone run, validate without regenerating unrelated artifacts:

```bash
GAIA_OPERATOR_OVERRIDE=1 gaia dev validate
```

For multiple rows, hand back to `/gaia-ingest-batch`, which performs exactly
one `gaia dev build` and final validation.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 323kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall