[lukemaj]

What: Separate the carried-edit registry from the app map and remove stale merged PR citations.
Why: Carried-file ownership and a GitHub-backed app map answer different questions, and merged secondary PR refs add stale bookkeeping.
So what: This is PR-A for #374; PR-B stacks on this branch, with independent reviews owned by the planner before merge.

Requirements and who asked: The maintainer's agreed council contract in #374 requests two sequential PRs and citation equivalence proof.
Deleted: Mandatory per-PR map accounting, merged secondary tokens, and registry follow-ups that did not change the carried file set.
Bottleneck: Prove deleting citations preserves map identity and folded GitHub states before adding Live parts.
Checked myself: Read the parser/fold and registry entries, checked #309 with git show --stat, and compared both maps over one live GitHub fact snapshot.

Refs #374.

Retitle the registry, cross-link the fork guides, narrow prs, validate URL shapes in the existing field loop, and test real headings/links plus rejection of an Issue URL in prs. #309 stays because it introduced the retained Vitest patch and cache test.

The starting map had more merged tokens than the contract's count of nine. Disposable /tmp/374-equivalence.mjs passed for 27 removed tokens, with 22 identical part IDs and primary refs and identical states over one live snapshot of 50 original refs. Kept %% BL: #289, PR #335: its comma makes #289 unreadable, so removing PR #335 fails primary-ref equivalence. All Issue refs remain.

Proof:

• node scripts/fork-features.mjs check: fork-features: OK (42 features).
• vp test run scripts/fork-maintenance.test.ts: 18 tests passed.
• Targeted oxfmt and oxlint passed; git diff --check passed.
• Fork-stack result will follow after checking the committed candidate against freshly fetched upstream.

No browser or dev server was used. Independent review remains for the planner. No versions changed, per repository policy.

Implementation: GPT-6.1-Sol, medium reasoning, Codex harness in T3 Code.

[lukemaj]

PR-A candidate 0c68c5550652c9c2e39617133a6ab361d844fdb3 passes bash scripts/fork-check.sh --base f798cb299ac9eab9d51ef63b5159de14a2b6aa18: fork-features: OK (42 features). and fork-check: OK. This is the real current merge base with freshly fetched upstream main, matching the script's CI base calculation. The initial check used a stale local upstream ref; fetching current upstream resolved that mismatch without source changes.

Targeted maintenance tests: 18 passed. Targeted formatter/linter and diff whitespace check passed. The live-fact citation equivalence result is in the PR body and #374. Independent reviews remain for the planner; nothing has been merged or deployed.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	20.9 KiB	20.9 KiB	0 B (0.0%)	29.3 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	+7 B (+0.1%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	+7 B (+0.6%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.2 KiB	21.2 KiB	0 B (0.0%)	29.3 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 3caa62c · PR result: 0c68c55 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.