Post-v0.18.0 work plan: estate bump, #107 validation wave, harness alignment, then Phase 2

[mmcky]

Tracking issue for the work queued behind today's release. Context: v0.18.0 shipped the forward --github integrity cluster (#108 → #104, #105, the mechanical halves of #106), criterion-score validation (#110 → #102), and node24 + @actions/* CJS majors (#111 → most of #89), all smoke-tested live on QuantEcon/test-translation-sync.fa#78. v0, v0.18, and v0.18.0 all point at the release. The items below are ordered; each is independently completable.

1. Bump the estate to v0.18.0

The reviewer resync fallback and all #108 fixes only reach target repos when their pins move — the estate sits at v0.16.1 (two releases behind; 9 pins across 4 repos, catalogue in QuantEcon/project-translation#7). Repos wired to @v0 picked the release up automatically; explicit ref: pins need PRs. While touching each repo's workflows, apply the current v0.17.0+ review-template guards (labeled trigger, concurrency group, permissions) where missing — the same drift #109 documents for the harness.

• lecture-python-programming (source workflows)
• lecture-python-programming.zh-cn
• lecture-python-programming.fa
• lecture-python-programming.fr
• intro.zh-cn wiring (per Phase 0 — wire sync automation into the two unwired zh-cn series (tracking) #74 / lecture-intro.zh-cn#212) if pinned explicitly

2. #107 validation wave (prompt change is live but unmeasured)

Re-run translate forward -f on two or three of the 11 hand-fixed intro.zh-cn lectures — long_run_growth, inequality, heavy_tails are the localisation-richest — and diff the output against the hand-fixed branches (the ground truth), counting CJK-bearing code lines per file as #107's follow-up comments prescribe (per-file counts catch wholesale anglicisation; exact-line diff catches the single load-bearing lines like long_run_growth's CSV mapping). Close #107 if preservation holds; either way the comparison seeds the #82 eval harness with its first real metric.

3. #109 harness alignment + the resync-review-label decision

The config work is listed in #109 (labels, current templates, status --write-state bootstrap, config-preserving resets). It contains one design decision to make deliberately: forward resync PRs carry action-translation-sync + resync but not action-translation, so review never auto-triggers on them anywhere. Recommendation: have createForwardPR also apply action-translation — no target-repo workflow edits needed, and it composes with #103's future verdict routing. Small related fix from the smoke: a missing label makes gh pr create fail after the branch is pushed (branch exists, no PR) — fall back to creating the label or opening the PR unlabeled with a warning.

4. Resume the #94 plan: Phase 2, then the #90 remainder

Unchanged from the plan and now unblocked: fence-aware parseSections (the intro_supply_demand phantom SOURCE_AHEAD on 2026-07-18 is a live false positive from this gap), #65 (label)= anchors, the round-trip invariant test, and a real validateMyST — then canonicalizeHeading() and fail-closed fetching from #90. The detection-beats-repair argument is why this outranks patching #90's defects individually.

Cheap wins, any time

• Glossary edits fail CI on a dist-action/glossary copy that nothing reads #95 — dist-action/glossary CI failure on a copy nothing reads
• docs: heading-maps.md documents a key format the action has never written #91 — heading-maps.md documents a key format never written (settle alongside canonicalizeHeading)
• Design: should French typography apply to the action's sync path, and how? #81 — likely closeable: v0.17.0 shipped typography in the sync path (fix(sync): apply typography in the sync path, typeset heading map with body #99); verify and close
• Work plan: suggested order for the post-review backlog #94 Phase 8 gardening — ~10 stale issues with pre-written close rationales

Deferred remainders (tracked in their own issues)

• node24: GitHub is already force-running the action on Node 24 (+ the @actions/* majors that clear the last advisories) #89: the ESM-only @actions/core 3.x / @actions/github 9.x migration (needs a Jest ESM story) and the optional @anthropic-ai/sdk bump.
• forward discovery misses content-stale files on unbootstrapped repos (found 12 of 41), and the summary miscounts what it did #106: --from-status / shared triage cache so forward can consume a check-sync run instead of paying for triage twice.
• forward resync reverts target-side localisation code — zh plots regress to English country names #107 stays open pending item 2's wave.

Refs: #94 (the standing work plan this extends), #108 #110 #111 #112 (today's PRs), #109 (harness), v0.18.0 release notes.

🤖 Generated with Claude Code

[mmcky]

Closing as superseded — twice over. #120 (2026-07-20) explicitly replaced this plan's sequence, and project-translation's PLAN.md is now the working plan of record across the engine and the estate.

Item disposition: estate bump — done and closed (project-translation#9, 9/9 pins on @v0, verified live 2026-07-21). #107 validation wave — still pending, carried in #120 P1 and PLAN Stage 3, unchanged in content. Harness alignment (#109) — substantially executed 2026-07-21 (targets aligned to production shape, @v0 pins, PAT steady state) with the remaining source-repo drift tracked in #109 itself. Resync-review-label decision — subsumed by the same harness/production alignment thread. #94 Phase 2 — in progress: item 1 (structural parity guard) shipped as v0.20.0; the round-trip invariant is next.

Nothing in this issue's content is lost — only its sequencing role, which #120 and PLAN.md now hold.