fix: repair OpenClaw skill YAML parsing

[Damin-Lee]

Summary

Fix three native OpenClaw SKILL.md files whose YAML frontmatter descriptions were invalid plain scalars, and add a regression test so the same parsing break does not come back. Also sync package.json with the current VERSION file so the existing version-consistency test passes again.

---

Type of Change

• Feature - New functionality
• Bug Fix - Bug fix
• Refactor - Code improvement without behavior change
• Documentation - Docs, comments, README updates
• Style - CSS/UI changes or code formatting
• Performance - Performance improvements
• Test - Test additions or modifications
• Chore - Build, config, dependency updates

---

Related Issues

• Closes #N/A
• Related to #N/A

---

Root Cause / Context

• Root cause: three hand-authored OpenClaw skills used plain description: scalars containing : , which is invalid YAML and caused Codex/OpenClaw skill loading to skip them.
• Why this approach: converting only the affected descriptions to block scalars is the smallest behavior-preserving fix. The added regression test guards the exact failure mode without forcing unrelated skills into block syntax.

---

What Changed?

Added

• A regression test in test/gen-skill-docs.test.ts that rejects native OpenClaw descriptions which would be invalid plain-scalar YAML.

Changed

• Converted frontmatter descriptions in:
  • openclaw/skills/gstack-openclaw-investigate/SKILL.md
  • openclaw/skills/gstack-openclaw-office-hours/SKILL.md
  • openclaw/skills/gstack-openclaw-ceo-review/SKILL.md
• Synced package.json version to VERSION.

Removed

• None

Fixed

• The native OpenClaw skill-loading failure that reported skipped skills due to invalid YAML.
• The existing package.json vs VERSION mismatch test failure.

---

Verification

• bun run check
• bun test
• bun run build
• Manual verification completed
• Not applicable items were explicitly marked in this PR body

Verification notes:

• bun run check: N/A, no check script exists in this package.
• bun test: passed for bun test test/gen-skill-docs.test.ts test/skill-validation.test.ts.
• bun run build: N/A for this PR.
• Manual verification:
  • ruby -e 'require "yaml"; ...' successfully parsed all native OpenClaw skills.
  • bun run skill:check passed, including OpenClaw skill inventory/freshness.
  • Full bun test still reports pre-existing golden baseline drift in test/host-config.test.ts for ship skill fixtures, unrelated to this diff.

---

How to Test

1. Open any of the three modified OpenClaw skill files and confirm the frontmatter uses description: |.
2. Run ruby -e 'require "yaml"; Dir["openclaw/skills/*/SKILL.md"].each { |f| YAML.load_file(f); puts f }'.
3. Run bun run skill:check and bun test test/gen-skill-docs.test.ts test/skill-validation.test.ts.

---

Screenshots

• N/A

PC (4K — 3840×2160)

N/A

Mobile (375px)

N/A

---

Documentation Updates

• README.md updated (for config/feature changes)
• Inline code comments added for complex logic
• API documentation updated (if applicable)
• N/A - No documentation updates needed

---

Breaking Changes

• No breaking changes
• Breaking changes (explain below):

N/A

---

Review Focus Areas

• Confirm the regression test catches only invalid plain-scalar YAML and does not overconstrain other native OpenClaw skills.
• Confirm syncing package.json to VERSION is the intended source-of-truth direction.

---

Dependencies

• No new dependencies
• New dependencies added:
  • Package:
  • Reason:

---

Deployment Notes

• No special deployment steps
• Environment variable updates required
• Database migration required
• Manual testing required in staging/production
• Other:

---

Additional Context

If a section does not apply, write N/A instead of deleting it.

This repo currently has unrelated full-suite failures in golden ship-skill fixture tests. They were present outside the scope of the OpenClaw YAML fix, so this PR keeps scope tight and fixes the broken skill-loading path plus its direct regression coverage.

---

Reviewer Checklist

• Follows project conventions
• Changes are appropriately tested
• Documentation is sufficient
• No unnecessary complexity
• PR size is appropriate (< 500 lines recommended)
• Commit messages are clear and descriptive