chore: release v0.29.2

[mmcky]

Release PR for v0.29.2 — a patch. Promotes [Unreleased] → [0.29.2] - 2026-09-21, bumps package.json / package-lock.json, and records the cut in .dev/STATE.md and the session log. No source change: dist-action/ rebuilds byte-identical.

What ships

Two payloads reach users for the first time here, because v0.29.1 was tagged and never released (its §4a gate came back 83/84; tally on #316; the CHANGELOG marks it [YANKED]):

• ml: stop the exercise rule priming invented directives — v0.29.1's gate failed scenario 17 (latent since v0.28.0) #317 (62504c1) — the repair that gate called for. The ml exercise-verbatim rule now says it concerns only directives the source already contains, so the model stops wrapping a plain ## Exercises list in {exercise-start}. Refusals by the structural-parity guard on the failing fixture: 5/12 at v0.29.0 (latent since v0.28.0), 11/12 at v0.29.1, 0/24 with the repair; local bundle through the action path on the failing fixture PR delivered 2 of 2.
• ml: encode the editor's round-3 answers — rules 27 → 28, glossary v0.7.0 (two shipped guesses corrected) #315 (cd41558) — the Malayalam editor's answers to the round-3 questions: ml rules 27 → 28, glossary v0.7.0, the corrected -ഉം comma rule, the plural rule, the Further Reading boundary record. Listed under [0.29.1] in the CHANGELOG, where the tag puts it.

Only the ml lane's prompt changes; zh-cn, fa and fr are untouched.

Checklist (AGENTS.md, Release Checklist)

• 1 — CHANGELOG promoted, fresh empty [Unreleased] above it; [0.29.1] stays, marked [YANKED]
• 2 — version bumped (package.json, package-lock.json)
• 3 — suite 1590 green; lint, format:check, check-dev-refs clean
• 4 — npm run build: dist-action/ unchanged
• 4a — after merge: tag v0.29.2, E2E-gate the tag. Read scenario 17 on the .ml lane first; then per-PR verdict coverage at engineVersion: 0.29.2. One draw per scenario is what the gate takes — the twelve-draw measurement in ml: stop the exercise rule priming invented directives — v0.29.1's gate failed scenario 17 (latent since v0.28.0) #317 is the evidence the defect is gone, the gate is the evidence the tagged bytes carry the repair
• 5 — move v0.29 and v0; @v0 smoke (scenario 01)
• 6 — GitHub release, title = the tag alone; notes cover both payloads

Gate tally and smoke read-off will be recorded here as comments.

🤖 Generated with Claude Code

[Copilot]

Copilot review overview

🟢 Approval recommended

The changes are consistent release metadata updates (versions, changelog, and internal logs) with no functional code modifications.

Review effort: Lite
Findings: None

What changed in this PR

Release bookkeeping for v0.29.2: bumps the npm package version, promotes the CHANGELOG entry for the release (including the v0.29.1 yanked note), and records the release cut in .dev/STATE.md and the session log.

Changes:

• Bump version 0.29.1 → 0.29.2 in package.json and package-lock.json.
• Promote CHANGELOG [Unreleased] → [0.29.2] - 2026-09-21 and retain [0.29.1] as [YANKED] with an explanatory note.
• Update .dev/STATE.md and .dev/log/... to reflect the v0.29.2 release cut and next steps.

File	Description
package.json	Updates package version to 0.29.2.
package-lock.json	Updates lockfile version fields to 0.29.2.
CHANGELOG.md	Adds 0.29.2 release section and marks 0.29.1 as yanked with rationale.
.dev/​STATE.md	Records v0.29.2 release cut in current project state.
.dev/​log/​2026-09-21-v0291-gate-scenario17.md	Appends note that #317 merged and v0.29.2 release PR was opened.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[mmcky]

§4a gate on the v0.29.2 tag (5f74d74): passed. Released.

	zh-cn	fa	ml
sync runs succeeded	28 / 28	28 / 28	28 / 28
translation PRs delivered	28	28	28
verdicts at engineVersion: 0.29.2	28 / 28	28 / 28	28 / 28

Census 9/9 workflows at @v0.29.2; 28 source PRs; 84/84 sync runs, none failed. Scenario 17 (new-document-toc) on .ml — the one v0.29.1's gate failed twice — delivered game-theory.md, its state file and _toc.yml (test-translation-sync.ml#242). One draw, as the gate takes; the twelve-draw measurement in #317 is the evidence on the rate.

One harness observation: the reset left one PR from the v0.29.1 gate open on the .ml lane (test-translation-sync.ml#192, stamped 0.29.1), so the lane briefly read 29 PRs / 28 stamped. Closed by hand; the script's "close all open PRs" step missed it, which is worth a look.

Checklist 5: v0.29 and v0 moved a6fda54 → 5f74d74. @v0 smoke (scenario 01): 3/3 sync runs succeeded; verdicts report engineVersion: 0.29.2, engineRef: v0 on all three lanes (test-translation-sync.zh-cn#962, .fa#346, .ml#251). Checklist 6: GitHub release published, title = the tag.