Incremental Change Surgeon

The problem

AI coding agents frequently rewrite entire files for small changes, leading to regressions, lost git history, and long review times. You end up spending more time debugging accidental side effects than the original task required.

What it does

• Enforces a "surgeon, not bulldozer" decision framework for every code modification.
• Prioritizes patch operations and targeted find-and-replace over destructive write_file commands.
• Calculates "blast radius" to identify all dependencies before executing changes.
• Mandates exact string matching and context-aware edits to preserve unrelated logic and formatting.
• Verifies minimal impact through mandatory diff audits and test suite execution.

Frameworks & tools

Universal compatibility with any language or framework. Optimizes workflows for Claude Code, Cursor, and Codex CLI.

Why this beats prompting it yourself

Standard prompts often fail to curb the LLM's bias toward regeneration when complexity increases. This skill installs a persistent multi-step verification loop that forces the agent to justify large edits and confirms no unrelated lines were touched before finishing.

Use cases

• Fixing specific bugs without affecting surrounding module logic.
• Adding new functions to large legacy files while maintaining existing git blame data.
• Updating configuration files like package.json or settings.py without reformatting the entire file.
• Refactoring internal function logic while strictly preserving signatures and call-site behavior.

Known limitations

This skill is not intended for generating new files from scratch, it only governs modifications to existing codebases.