[wukongnotnull]

This PR MUST target the main branch. Active work is developed on
focused branches and merged directly into main after review.

Tracking: #56

Who is submitting this PR? (required)

Field	Value
Your model + version	Cursor Grok 4.7
Harness + version	Cursor cloud agent
All plugins installed	none beyond the Cursor cloud-agent environment (MCP namespaces present: cursor-cloud, cursor-subscriptions, Gmail; cursor-cloud was used only to read this run's model id; the others were unused)
Human partner who reviewed this diff	悟空非空也g — asked to 发版. origin/main already contains merged #55

What problem are you trying to solve?

Manifests on main still say 6.5.0 while PR #55 is already merged. Users who install by version, and bug reports that cite 6.5.0, cannot distinguish the grilling HANDOFF change from the tagged release. RELEASE-NOTES.md has no dated section for that commit. 悟空非空也g asked to 发版 on current origin/main (tag v6.5.0 plus #55).

This is the same problem #54 solved for 6.5.0 and #39 solved for 6.4.0: ship the version number and notes together so harness install paths and the changelog match main.

What does this PR change?

Runs scripts/bump-version.sh 6.6.0 on the seven declared files (package.json, Claude/Cursor/Codex/Kimi plugin manifests, Claude marketplace plugin version, product-design.lock.json wukong_code_version). Adds ## v6.6.0 (2026-10-14) describing only PR #55. Updates the README Product Design version line and the docs/prd.md current-version line so the snapshot matches the tag.

Is this change appropriate for the core library?

Yes. This is core plugin release metadata and notes, not a new skill, not a third-party dependency, and not a rewrite of behavior-shaping skill text. The grilling text already landed in #55. The lock-file wukong_code_version field is outside the hashed Product Design import boundary (same as #39 and #54).

What alternatives did you consider?

• Patch 6.5.1. Rejected: feat: grilling HANDOFF writes a derived technical spec #55 is a user-visible skill behavior change (feat: derived technical spec instead of persisting the Shared-Understanding Record as the spec), not a broken-install fix or a docs-only typo. Same reasoning as Release v6.5.0 #54 rejecting 6.4.1 when Unreleased already held a user-visible feature.
• Major 7.0.0. Rejected: no harness install incompatibility or removals. A skill handoff change is not a major on this repo (the previous grilling handoff shipped in 6.4.0).
• Leave feat: grilling HANDOFF writes a derived technical spec #55 out of the notes. Rejected: it is the only commit range after v6.5.0 (1b3542e, 840317c, merge 3b81a61).
• Invent extra features in the notes. Rejected by the release request.
• npm publish or a Codex portal zip in this PR. Out of scope. registry.npmjs.org/wukong-code is 404 (re-checked this session). The v6.5.0 GitHub Release has no assets. scripts/package-codex-plugin.sh is maintainer-side and was not attached to Release v6.5.0 #54's release. After merge, the follow-up is an annotated v6.6.0 tag and a GitHub Release whose body is the dated notes.

Does this PR contain multiple unrelated changes?

No. Every file is the version bump, the release-notes heading for that bump, or the two version strings the previous release PR also updated (README Product Design line and docs/prd.md).

Existing PRs

• I have reviewed all open AND closed PRs for duplicates or prior art
• Related PRs: Release v6.5.0 #54 (Release v6.5.0 — same path; this is the next cut, not a redo), Release v6.4.0 #39 (Release v6.4.0), docs: add Unreleased release-notes entry for #37–#45 #46 (Unreleased notes, already consumed by Release v6.5.0 #54), Release v6.2.0 #4 / Release v6.2.1 #10 (older release PRs), feat: grilling HANDOFF writes a derived technical spec #55 (content being dated, not re-landed). No open duplicate release PR. No open or closed PR titled for 6.6.0.

Difference from #54: that PR dated everything after 6.4.0. This PR dates only #55. It does not change skills or tests.

Environment tested

Harness (e.g. Claude Code, Cursor)	Harness version	Model	Model version/ID
Cursor cloud agent (this checkout)	Cursor cloud agent Linux VM, Oct 2026	Cursor Grok 4.7	grok-4.7

Local: bash scripts/bump-version.sh --check — all seven files at 6.6.0; audit found no undeclared 6.6.0 strings. bash tests/skills/test-gemini-retirement.sh — PASS including ## v6.6.0 in release notes. npm run test:extended — PASS after installing host rsync (preflight needs it; not a new plugin dependency). shellcheck 0.9.0 is installed for the CI lint job; this PR does not change shell scripts. CI on this PR is the extended suite plus ShellCheck.

New harness support (required if this PR adds a new harness)

Not applicable. This PR does not add a harness.

Clean-session transcript for "Let's make a react todo list"

N/A — not a new harness integration.

Evaluation

• What was the initial prompt you (or your human partner) used to start the session that led to this change? 悟空非空也g asked to 发版 on current origin/main, which includes v6.5.0 and merged feat: grilling HANDOFF writes a derived technical spec #55.
• How many eval sessions did you run AFTER making the change? Zero skill-behavior evals. This PR does not change skill or hook text. feat: grilling HANDOFF writes a derived technical spec #55 already recorded the grilling RED/GREEN probes.
• How did outcomes change compared to before the change? Version check and the Gemini retirement notes gate now see 6.6.0; the extended suite still passes. No new agent-behavior claim.

Language-pack evidence

• Language and support status: n/a (no language pack change)
• Real failure generic methodology handled incorrectly: n/a
• Project markers and toolchain versions: n/a
• Official sources: n/a
• No-guidance repetitions and failures: n/a
• With-guidance repetitions and results: n/a
• Monorepo, unsupported-language, and docs-only controls: n/a
• Human reviewer familiar with the language: n/a
• Frameworks or third-party preferences introduced: none

Rigor

• If this is a skills change: I used wukong-code:writing-skills and completed adversarial pressure testing (paste results below)
• This change was tested adversarially, not just on the happy path
• I did not modify carefully-tuned content (Red Flags table, rationalizations, "human partner" language) without extensive evals showing the change is an improvement

Not a skills change. Adversarial here means: bump-version.sh --audit reports no undeclared 6.6.0 hits (template lockfiles still contain unrelated fdir@6.5.0, which is why they stay out of .version-bump.json); the Gemini retirement test requires the dated ## v6.6.0 heading; the extended suite was run, not only --check.

Human review

• A human has reviewed the COMPLETE proposed diff before submission

悟空非空也g explicitly asked to 发版 this repo, authorized a release PR targeting main, and #55 (the only post-6.5.0 product change) is already merged. The diff in this PR is version metadata and notes for that already-merged history. The human should still glance at RELEASE-NOTES.md wording on GitHub before relying on the GitHub Release object.