The Agent Skills Directory

[SAFE]: The skill implements a robust refactoring process centered on behavior preservation. It establishes a test baseline before changes, uses characterization tests for gaps, and verifies every incremental step against the full test suite.
[COMMAND_EXECUTION]: The skill executes project-specific commands for testing, linting, and type-checking. These commands are discovered from local project files (e.g., package.json, Makefile) and are necessary for the skill's primary purpose of verifying refactoring safety.
[INDIRECT_PROMPT_INJECTION]: The skill processes existing source code and project configuration files as input. While this represents a potential attack surface for indirect prompt injection, the risk is inherent to the tool's primary purpose of code analysis and is mitigated by the mandatory verification of changes against the existing test suite and quality gates.
Ingestion points: The skill reads source code, CLAUDE.md, and project configuration files (e.g., package.json, pyproject.toml) to map dependencies and discover build commands in Phase 0 and Phase 1.
Boundary markers: The skill does not use specific delimiters or "ignore" instructions when reading source code for analysis.
Capability inventory: The skill can execute shell commands for testing, linting, and type-checking, and can write changes to files and create git commits.
Sanitization: No explicit sanitization of ingested code content is performed.

refactor