What does this skill do?

The CodeHealth MCP Skill provides structured feedback on code maintainability during AI-assisted editing. It complements lint and style tools with design-level health scores and regression gates, so that each change can be verified before being committed via CodeScene’s MCP server.

Pre-publication review
Perform a complete structural analysis before modifying a hotspot, legacy module, or unknown file.
Delta Verification
Get a quick score after each change and compare it to the baseline to detect regressions.
Commit Backup
Block commits that introduce code health regressions before they reach the repository.
Branch Analysis for PR
Check the set of changes at the branch level before opening a pull request against the main branch.

Usage examples

🔍 Preliminary Review
Run `code_health_review` on `src/payments/processor.ts` before making any changes, and let me know the baseline score and the code smells that were detected.
📊 Delta Verification
I just refactored the authentication module. Run `code_health_score` and compare it to the baseline score of 6.3.
🛡️ Commit Backup
Before committing these changes, run `pre_commit_code_health_safeguard` and block the commit if there are any regressions.
🔀 PR Analysis
I'm going to open a PR. Run `analyze_change_set` against the `main` branch to verify that the code's health hasn't deteriorated.

Features

Comprehensive Structural Analysis code_health_review provides a design-level analysis that identifies code smells and offers recommendations before you make any changes.
Quick Delta Score code_health_score returns an instantaneous numerical score to compare before and after each change.
Regression gate in commits pre_commit_code_health_safeguard blocks commits that introduce maintainability regressions.
Branch-Level Verification analyze_change_set summarizes the impact of the change set before opening a pull request.
Health Scale 1–10 Use green (9–10), yellow (4–8.9), and red (1–3.9) scores to adjust the scope of each edition.

Frequently asked questions

No. The skill works with CodeScene's free standalone token. You just need to set CS_ACCESS_TOKEN in your MCP block.
The skill does not generate scores. It informs the user that the check has been skipped and proceeds only with explicit approval, prioritizing lint and tests as an alternative path.
The scale ranges from 1 to 10: 9–10 is green (healthy), 4–8.9 is yellow (technical debt; proceed with caution), and 1–3.9 is red (severe debt; make only minimal changes).
Yes. You can add the compliance block to AGENTS.md or CLAUDE.md so that Claude automatically runs the tools before commits and PRs.
CodeHealth MCP — Code Health Analysis with Claude AI

¿Prefieres escuchar el contenido? Genera la narración de audio con un clic.