feat(parity): migrate fork manifest to lifecycle schema v2

[Kyzcreig]

Worker t_12df7dc2's schema-v2 migration for docs/sync/fork-features.json: per-entry lifecycle classifications + lint contract in scripts/hermes_parity/lint_manifest.py.

Verified fresh: 33/33 parity tests green in an isolated worktree. Landing this FIRST unblocks the absorption-sweep commit (t_d82545b8) stacked on it — the sweep flips entries to absorbed/in-flight based on today's 19-PR upstreaming wave.

Reviewed per the worker-handback gate; kanban cards close on merge.

[Kyzcreig]

FleetReview

Confidence: 3/5

Findings

• P1 scripts/hermes_parity/lint_manifest.py:59 — Schema Bypass
• P1 scripts/hermes_parity/forkdelta.py:54 — Type Erasure
• P1 scripts/hermes_parity/forkdelta.py:53 — Missing lifecycle fields bypass the documented rollout warning

---

FleetReview provenance · models: B=gpt-5.6-sol, C=claude-code-opus-4-8, F=gpt-5.6-sol, G=grok-4.5 · cost: $4.48 · duration: 12m 51s · rounds: 1 · files examined: 4

[Kyzcreig]

Closing as superseded — the schema-v2 migration is already on main, landed via the stacked sweep PR #436 (0b7cc2b6c).

Ground-truth verification (2026-08-06):

Manifest structure is identical — fork/main's docs/sync/fork-features.json and this branch's version carry the same key set:
['absorbed_date','feature','lifecycle','paths','tests','upstream_ref','why']

Entry counts: main 12 vs this branch 11, and main has more classified entries (absorbed: 2 vs 1). Main is strictly ahead.

Code half also present on main:

• scripts/hermes_parity/forkdelta.py:15 — LIFECYCLES = frozenset({"upstream-intended", "fork-permanent", "absorbed"})
• scripts/hermes_parity/lint_manifest.py:56 — def lint_schema(...)
• lint_manifest.py:68 — the absorbed feature requires upstream_ref guard
• tests/scripts/test_hermes_parity.py — 8 lifecycle/lint_schema references

Why it read as DIRTY/CONFLICTING: the branch is 2 commits ahead but 35 behind main. Test-merging onto current fork/main conflicts solely in docs/sync/fork-features.json (6 conflict markers) — this branch would re-apply an old-schema→v2 migration to a file that is already v2 and has since gained entries. Rebasing would mean resolving a migration against its own completed output.

No work is lost: the PR body's stated purpose ("landing this FIRST unblocks the absorption-sweep commit stacked on it") was satisfied when the sweep itself landed and carried the schema with it. Classic stacked-PR closeout — a squash merge of the tip does not auto-close contained siblings.