| name | hima-stewardship |
| description | Ongoing post-ship guardianship — dependency hygiene, regression watch, dead-code sweep, health scorecard against real code. |
| stage | maintenance |
| source | base |
| mode | advisory |
Purpose
Keep a shipped system healthy after the cycle closes. hima-stewardship is the maintenance-stage
skill: it scans the actual codebase (not aspirational docs) for drift, decay, and risk that
accumulates between cycles — dependency staleness, dead code, regression exposure, and rising
complexity trends.
When to use
- Scheduled/periodic sweep after a feature ships, or monthly/quarterly per the project's
discipline cadence — not a per-commit gate.
- Advisory by default: it reports findings, it does not block a commit or a stage transition.
Steps
- Scan for dead code, unused exports, and duplicate logic (e.g.
knip-style sweep).
- Check dependency staleness and known vulnerabilities (
npm audit, npm outdated).
- Re-check codebase health metrics against target thresholds (complexity, lint warnings, type
coverage, build time) and flag rising trends as refactor signals.
- Verify prior verdicts still hold — no silent regression on previously DONE_VERIFIED work.
- Produce a scorecard grounded in real code state, not the last spec's claims.
Done when
A scorecard is produced comparing real code state to health targets, findings are named
explicitly (not silently absorbed), and any regression against a prior verdict is flagged for
a follow-up hima-triage or hima-forge pass.
