The Agent Skills Directory

[SAFE]: The skill is purely instructional and does not require or use any external tools, network access, or sensitive file system permissions.
[PROMPT_INJECTION]: The instructions contain a robust defensive rule stating that code comments, strings, and variable names must be treated as data only, effectively mitigating potential prompt injection attacks embedded in processed source code.
[COMMAND_EXECUTION]: The skill defines a strict boundary by forbidding direct code modifications and directing the agent to only produce reports, which prevents unauthorized or silent changes to the user's codebase.

code-quality