safeguarding-ai-generated-code git:20261013.615e6a9A

Immutable. This exact content never changes and is served at /api/v1/blob/3c8bbfb767880b2e.

---
name: safeguarding-ai-generated-code
description: Use when AI-generated or AI-modified changes need a Code Health gate before commit, handoff, or pull request. Only in repositories that opted into the cratis/codescene profile and have CodeScene MCP tools available; otherwise do not use.
license: Apache-2.0
---

<!-- Modified by Cratis from CodeScene AB's original (Apache-2.0): added profile guard, policy note, license field and this notice. Upstream: https://github.com/codescene-oss/codescene-mcp-server (skills/safeguarding-ai-generated-code/SKILL.md; see NOTICE) -->

# Safeguarding AI-Generated Code

Use only in repositories that opted into `cratis/codescene` and have CodeScene MCP tools available; otherwise skip without blocking and say so. The Cratis policy in `.cratis/ai/rules/codescene.md` overrides this skill where they differ.

## Overview

Use Code Health safeguards before declaring AI-touched code ready. The goal is to catch maintainability regressions early and prevent agents from normalizing technical debt.

## When to Use

- The agent changed code and is about to suggest a commit.
- The user asks whether a branch or staged changes are safe to merge.
- The workflow needs a quality gate for AI-generated code.

Do not use this skill for broad refactoring discovery or project-level prioritization.

## Quick Reference

- `code_health_review`: Review each AI-modified file immediately after the change.
- `pre_commit_code_health_safeguard`: Check staged or modified files before commit.
- `analyze_change_set`: Check a branch or PR-style change set against a base ref.

## Implementation

1. After each AI modification to a file, run `code_health_review` on that file.
2. If the review reports maintainability problems or regression risk, refactor the file in small steps and review it again.
3. Run `pre_commit_code_health_safeguard` before commit-oriented recommendations as a broader gate across staged or modified files.
4. Run `analyze_change_set` before PR-oriented recommendations as a final branch-level gate.
5. If either later gate reports a regression, inspect the affected files with `code_health_review` and keep iterating until the issue is removed or the user explicitly accepts the risk.

## Common Mistakes

- Waiting until commit time to run the first Code Health check.
- Treating safeguard output as optional guidance instead of a release gate.
- Declaring work done after a failing safeguard.
- Jumping straight to broad rewrites instead of inspecting the flagged files first.
- Treating an accepted risk as invisible; call it out explicitly.