The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/de_ai_audit.py dynamically loads and executes an external script prose-audit.py using importlib.util.spec_from_file_location and loader.exec_module. The path is computed dynamically (Path(__file__).resolve().parents[3] / "scripts" / "prose-audit.py"), reaching three levels above the script's directory to find a shared dependency. This technique of executing code from computed paths is a security risk if the file system hierarchy is not strictly controlled.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest, analyze, and rewrite untrusted text provided by users to remove machine-writing patterns. 1. Ingestion points: Untrusted data enters the context through files read by the Read tool and the audit_file function in scripts/de_ai_audit.py. 2. Boundary markers: The skill instructions do not define specific delimiters or boundary markers to isolate user-supplied prose from the agent's operating logic, nor are there instructions to ignore instructions embedded within the text. 3. Capability inventory: The agent is granted permissions for Bash, Write, Edit, and Read, and is instructed to execute shell commands via uv run in its primary workflow. 4. Sanitization: There is no evidence of sanitization, escaping, or filtering of the external content before it is processed by the agent.

de-ai-revise