The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates by scanning and processing external codebase files using Read, Glob, and Grep tools. This data is then used by the AI agent to make decisions about code removal or modification. There is a theoretical risk that malicious instructions embedded within the analyzed code (e.g., in comments or string literals) could attempt to influence the agent's behavior.
Ingestion points: Source code files and configuration files read via the Read, Glob, and Grep tools.
Boundary markers: The skill does not explicitly define instructions to the LLM to ignore embedded commands within the code it analyzes.
Capability inventory: The skill has access to high-impact tools including Write, Edit, and Bash (restricted to specific commands like git, npm, and python3).
Sanitization: No explicit sanitization or filtering of codebase content is performed before the agent processes it.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file utilizes the dynamic context injection pattern (!command) to execute git, ls, and cat commands at skill load time. This is used to provide the agent with immediate context regarding the project structure and recent changes. These commands are hardcoded and benign in the current implementation.
[UNVERIFIABLE_DEPENDENCIES]: The skill instructions and documentation recommend the use of several third-party Node.js and Python tools via npx and standard package managers (e.g., knip, madge, jscpd, ruff, bandit). These are well-known developer tools and their use is consistent with the primary purpose of the skill.

cleanup-code