faff-tidy

Run a tidy pass that re-fetches every active issue with its comments, classifies each into buckets (needs action, ready, almost ready, parked, on hold), and applies mechanical fixes like stripping dead blocker links. Judgement calls such as cancellations are surfaced for human decision, never auto-applied.

The tidy pass flags duplicates covering the same work, stale specs whose design no longer matches the codebase, superseded specs whose premise was replaced by shipped work, and aging or dead-weight issues. Each finding comes with a recommended action such as merge, archive, or re-prep.

Yes, but only issues with a real discoverable spec, no open blockers, and automation eligibility are promoted. Issues with only a description are never promoted; they are flagged for the prep skill instead, and eligibility labels are always human-gated.

The diagnostics pass detects dependency cycles, ghost-project pointers to nonexistent containers, repeat-park patterns, splittable specs, and chain gaps where spec-implied work has no tracking ticket. Mechanical fixes like stripping defensive cycle edges are applied; the rest are surfaced.

No. Labelling is delegated to the prep skill, and tidy never suggests label changes. Cancelled work is ignored entirely, and descendants of cancelled ancestors are surfaced for human decision rather than auto-cancelled.