The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and process arbitrary frontend code from a user's project, creating a surface for indirect prompt injection where instructions embedded in the code could influence agent behavior.
Ingestion points: Processes project files based on the $ARGUMENTS target or the entire project scope.
Boundary markers: The skill lacks explicit delimiters or instructions to ignore instructions found within the processed data.
Capability inventory: The skill possesses the ability to read existing project files, modify them, and create new documentation files in the docs/ directory.
Sanitization: No mechanisms for sanitizing or filtering input from the processed files are specified.
[COMMAND_EXECUTION]: The skill uses dynamic context injection placeholders to retrieve system metadata when the skill is loaded.
Evidence: Use of !date and `!`node --version to inject the current timestamp and Node.js environment version.
Risk assessment: The executed commands are benign information-gathering tools and do not incorporate user-supplied parameters, minimizing the risk of arbitrary command injection.

enhance