feat: add missing agent personas, fix /lfg command, and cleanup CORA references

[terry-li-hm]

Summary

This PR combines several maintenance tasks and bug fixes into a single comprehensive update for the compound-engineering plugin:

1. New Agent Personas: Added 9 missing agent definition files (frontend-design, brainstorming, etc.) to match the plugin documentation.
2. LFG Command Fix: Resolved broken command references and Ralph-Wiggum namespace typos in the /lfg command for OpenCode users (fix: resolve legacy ralph-wiggum namespace typo in lfg command and converter #126).
3. CORA Cleanup: Removed remaining hardcoded "CORA" legacy references and renamed cora-test-reviewer to test-quality-reviewer (fix: remove remaining CORA references missed by #121 #127).
4. Versioning & Metadata:
   • Bumped version to 2.29.0 (Minor).
   • Updated component counts and agent tables in README.md.
   • Updated CHANGELOG.md with full version history.
   • Standardized all agent descriptions to the mandatory third-person format.

Impact

• The /lfg autonomous workflow is now fully functional for OpenCode users.
• Specialized agents now have their intended instructions/personas.
• Documentation and metadata are fully synchronized with the actual codebase state.

---

🤖 Generated with [Claude Code]

[terry-li-hm]

Hi @kieranklaassen, I've consolidated my recent bug fixes (LFG typos, CORA cleanup) and missing agent personas into this single PR to prevent versioning conflicts.

I've also done a full compliance pass to ensure all skill references use proper markdown links per CLAUDE.md and added a verify_plugin_metadata.py script to automate component counting. This PR is now fully synchronized and ready for a final review!

[kieranklaassen]

Thanks! Will check all these out over the weekend!

[terry-li-hm]

Closing this omnibus PR. The core change — adding agent personas for 9 skill-only components — I now believe is the wrong direction. See #156 for the reasoning:

These components (frontend-design, brainstorming, compound-docs, agent-browser, etc.) work as skills. Adding agent personas makes them eager-loaded into every session's Task tool description, adding token overhead for all users on every API call. Skills are lazy-loaded and near-zero cost when idle.

The better fix for workflows that try to spawn these as agents (e.g. /workflows:compound → compound-docs) is to update the workflow commands to use Skill invocations instead of Task agent spawns.

The /lfg and CORA fixes in this PR were already addressed by #121 and #126 (both merged). The only remaining value was the agent personas, which #156 argues against.

Apologies for the noise — my thinking evolved after living with the trim-based workaround and seeing the token impact firsthand.