canon: link-rot Phase 2 shipped — persist handoff + ledger + spec v2.2

[klappy]

Closes out the v0.26.0 oddkit promote — link-rot Phase 2 — by persisting the session artifacts and aligning the audit spec with what shipped.## What this PR adds**odd/handoffs/2026-04-27-link-rot-phase-2-promote-resume.md** — the resume handoff that this session executed. Authored by the prior session (compacted at 2026-04-27T02:39 UTC), kept Continuation Test discipline; this session executed it cleanly. The companion "original" handoff (2026-04-27-link-rot-phase-2-promote-and-phase-3.md) was attached to the prior session's chat context only and was lost — see the resume's "Persistence Note" appendix. Per axiom 4 ("You Cannot Verify What You Did Not Observe"), this PR refuses to reconstruct the original from inference. The resume is comprehensive enough to stand alone.odd/ledger/2026-04-27-link-rot-phase-2-shipped.md — session ledger documenting the v0.26.0 ship: timeline, decisions (lineSeen waiver, spec amendment, original-handoff loss), observations (DNS overflow workaround, Bugbot Autofix non-blocking carve-out, real link-rot in writings/), learnings, and what remains for Phase 3.docs/oddkit/specs/oddkit-audit.md v2.1 → v2.2 — defers index_state field and PARTIAL_INDEX status enum value per validator findings F-3 and F-5. Both were intentional drift in the v0.26.0 implementation; rather than carry permanent spec-vs-shipped drift, the spec now matches shipped behavior with both deferred items marked clearly in Partial-Data Compliance and the Origin v2.2 amendment paragraph. Use Only What Hurts: if a real consumer demonstrates need to distinguish "URI dead" from "couldn't be checked because index isn't warm," the deferred mechanism graduates back into the schema.F-4 (suppressed_findings returns full objects vs spec's "count only") was dispositioned by the validator as no action — additive, not harmful and remains unchanged in v2.2.## What shipped to prod (verification recap)- ✅ oddkit_version returns 0.26.0- ✅ Unified oddkit action enum at 13 values (audit + resolve + 11 prior); strictly +1 from v0.25.0 baseline- ✅ Standalone oddkit_audit tool live; default scope ["writings/"]; wider scope ["writings/", "canon/"] honored- ✅ Real link-rot surfaced: 3 dead-references in default-scope smoke; 17 in writings+canon scope- ✅ Cursor Bugbot finding on promote PR head dispositioned per release-validation-gate Rule 1 (waived in body + tracked at klappy/oddkit#148)- ✅ Validator findings table appended to klappy/oddkit#146 body before merge## Refs- Promote PR: klappy/oddkit#146 (merged 2026-04-27T02:51:55 UTC)- F-1 fix-forward: klappy/oddkit#147- Bugbot lineSeen tracking issue: klappy/oddkit#148- Resume handoff this PR persists: klappy://odd/handoffs/2026-04-27-link-rot-phase-2-promote-resume- Session ledger this PR adds: klappy://odd/ledger/2026-04-27-link-rot-phase-2-shipped- Spec at v2.2: klappy://docs/oddkit/specs/oddkit-audit- Canon honored: klappy://canon/constraints/release-validation-gate (Rules 1, 2, 3 all)## Next session pick-upPhase 3 PR-3.1 — .github/workflows/canon-quality.yml on this repo (klappy.dev), soft-block mode. Spec at v2.2 is the source of truth. Continuation Test in the new session ledger.

---

Note

Low Risk
Documentation-only changes that align the spec with the implemented oddkit_audit output and persist operational notes; no runtime behavior is modified.

Overview
Aligns the oddkit_audit action spec to shipped v0.26.0 by bumping to DRAFT v2.2 and removing index_state/PARTIAL_INDEX from the documented output schema (marking the partial-index mechanism as deferred).

Adds two canon artifacts: a promote-resume handoff for completing/validating the Phase 2 promote steps, and a session ledger capturing what shipped, validator dispositions, and remaining Phase 3 follow-ups.

^{Reviewed by Cursor Bugbot for commit 7908310. Bugbot is set up for automated code reviews on this repo. Configure here.}