The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided source code for refactoring analysis. This creates an attack surface where malicious instructions could be embedded in the code being analyzed to influence the agent's behavior.
Ingestion points: The skill processes user input via the $ARGUMENTS variable and analyzed code blocks in SKILL.md.
Boundary markers: The instructions do not specify strict delimiters for untrusted code, although they reference an external playbook for implementation patterns.
Capability inventory: The skill is scoped to analysis, planning, and applying code changes. It references the use of static analysis tools (Ruff, Mypy, ESLint, Semgrep, CodeQL) and refactoring utilities (Sourcery).
Sanitization: No specific sanitization or escaping mechanisms for the analyzed code content are described in the instructions.
[SAFE]: The skill instructions and resources focus exclusively on code quality, maintainability, and standard software engineering practices. The resource file provides standard configuration examples for well-known linting and security tools. The skill also includes a specific 'Behavior preservation and evidence' section that acts as a guardrail against hallucinating security metrics or executing unverified code snippets.

codebase-cleanup-refactor-clean