The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md file instructs the agent to execute a Python script located at scripts/invoke.py using the uv package manager. This is the intended primary execution path for the skill.
[DYNAMIC_EXECUTION]: The scripts/invoke.py script uses the subprocess.run method to execute a secondary engine script (kntnt.py). The script includes logic to dynamically locate this engine by searching through a predefined list of common AI agent and IDE configuration directories (e.g., .claude/skills, .cursor/skills, .roo/skills) in the user's home directory to ensure the tool works across different environments.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources such as URLs and local files provided by the user to generate drafts.
Ingestion points: The SKILL.md file explicitly mentions gathering material from local paths and URLs provided in the user's brief.
Boundary markers: No technical delimiters or 'ignore' instructions are used to wrap the ingested content, although the skill emphasizes 'Source Fidelity' as a primary instruction.
Capability inventory: The skill has access to command execution via uv run and subprocess.run in its associated scripts.
Sanitization: There is no evidence of input sanitization or filtering of the external content before it is passed to the drafting engine.

write