The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the kai-gate command-line tool to perform quality scoring and policy enforcement on markdown files.
[COMMAND_EXECUTION]: Interpolates file paths and policy names directly into shell commands (kai-gate score {file}). While standard for such tools, it relies on the execution environment to handle variable sanitization to prevent command injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted markdown content from files, creating an attack surface where embedded instructions could attempt to influence the scoring output or the auto-fix logic.
Ingestion points: Reads content from user-specified file paths or the ~/.kai-marketing/drafts/ directory.
Boundary markers: No explicit delimiters or 'ignore' instructions are used when passing the file content to the scoring utility.
Capability inventory: Subprocess execution of kai-gate and file-write access to ~/.kai-marketing/gates/.
Sanitization: Content is passed directly to the scoring utility and potentially an LLM (for 'Four U's' scoring) without explicit sanitization or filtering of embedded instructions.

content-gate