fix: align persistent stop hook and tighten agent output contracts

[Yeachan-Heo]

Summary

• point installed persistent-mode hook wiring at the maintained persistent-mode.mjs path
• add concise-output guidance to autopilot / ultrawork / ralph agent prompt surfaces
• add targeted regression coverage for runtime hook wiring and output-contract guidance

Testing

• npx vitest run src/hooks/autopilot/tests/prompts.test.ts src/hooks/autopilot/tests/pipeline.test.ts src/tests/ralph-prd-mandatory.test.ts src/tests/issue-2652-runtime-wiring-and-output-contract.test.ts
• npx vitest run src/hooks/persistent-mode/stop-hook-blocking.test.ts

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 89eb0de5ea

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Keep Stop hook on feature-parity implementation

Changing the Stop hook target to scripts/persistent-mode.mjs removes continuation safeguards for active Skill-tool workflows, because that script currently lacks the skill-active-state.json blocking path that exists in scripts/persistent-mode.cjs (the cjs implementation still enforces reinforcements before allowing stop). In sessions where skills like plan, code-review, or tdd are mid-flight, Stop can now pass through early and terminate the run before the skill finishes.

Useful? React with 👍 / 👎.