Proposal: codehealth-mcp - Code Health MCP skill for agent workflows

[adnasal]

Hi - sharing a skill proposal since PRs from forks look restricted to collaborators on this repo.

The idea

A skill that wires the CodeScene Code Health MCP into the same kind of disciplined loops your other skills teach (TDD, diagnose, etc.), but for maintainability:

• Review structure before editing
• Check the score after each change (no "done" if it dropped)
• Gate commits and PRs that worsen Code Health

Uses the free standalone token only - four MCP tools, no enterprise account.

Proof point

On Flask (pallets/flask), that loop reportedly moved Code Health 4.82 → 9.1 using only standalone tools.

Implementation (ready to review)

I've pushed a branch on my fork:

• Branch: https://github.com/adnasal/skills/tree/add-codehealth-mcp-skill
• Diff vs main: main...adnasal:add-codehealth-mcp-skill

Tracking issue: #260

Two files: new skills/codehealth-mcp/SKILL.md + one README table row.

Happy to iterate on feedback. Thanks!