Skip to content

gaia-research/gaia-skill-tree / .agents/skills/gaia-steward-observe

gaia-steward-observe

Verified

Observe the repository with every registered Gaia Steward detector and report what each one actually saw, without changing anything. Use for "/gaia-steward-full --observe", "what is Steward seeing", "scan the repo for maintenance debt", or as the first phase of a full Steward run. Distinguishes actively observed, unknown/unavailable and not-evaluated detectors; never treats an unexecuted probe as healthy.

gaia-steward-observe

Responsibility

Run the detector fleet once against the checkout head and produce a receipt that says, per detector, whether it observed, could not observe, or was not evaluated. Nothing else.

Input

  • A Git worktree of gaia-skill-tree (the detectors read Git's index, not the filesystem).
  • Optional driver inputs under the ignored .gaia/steward/inputs/: ci-runs.json, advisories.json, test-history.json. Their absence is a not-evaluated state, never an error and never healthy.
  • The GitHub pull-request census github-prs.json (routine 19, pr-maintenance), refreshed by python scripts/steward_pr_census.py. A missing census reports routine 19 not evaluated; an incomplete or stale one reports it unavailable (Class A is blocked).
  • The GitHub issue census github-issues.json (routine 18, issue-hygiene), refreshed by python scripts/steward_issue_census.py before observing. A missing census reports routine 18 not evaluated. An incomplete or stale census reports it unavailable, which blocks Class A for the run.

Output

gaia steward full run --observe --json
python scripts/check_steward_coverage.py --receipt .gaia/steward/receipts/<run>.json

The receipt (steward-receipt-v1) carries coverageKnown, coverageUnknown, notEvaluated, openDebt and authorityCounts. The checker turns it into a live state per detector: actively-observed only when listed in coverageKnown; everything else is unknown-unavailable.

Authority ceiling

Read-only. Writes only ignored state under .gaia/steward/ and the run record in the Steward ledger. No repair, dispatch, branch or PR.

Stop conditions

  • Any detector in coverageUnknown: report it first and stop treating the run as clean; Class A is blocked for the run.
  • The observation budget in POLICY.yaml is exceeded: stop and report.

Idempotency

Observing the same head twice yields the same debt ids (debt:{kind}:{slug}:{digest}); re-observation updates lastSeen, never duplicates debt.

Failure states

StateMeaning
coverageUnknowna required input was missing, truncated or unreadable
notEvaluatedan optional driver input was not supplied
blocked resultat least one detector was unknown

Validation

python scripts/check_steward_coverage.py
python -m pytest tests/steward/test_controller.py tests/steward/test_coverage_regressions.py -q

Tests

tests/steward/test_controller_resolution.py, tests/steward/test_routine_coverage.py, tests/steward/test_steward_skills.py.

slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall