The Agent Skills Directory

[DATA_EXPOSURE]: The script scripts/search_references.py reads local markdown files from the references/ directory. It includes a path validation check (_validate_refs_dir) using Path.resolve() to ensure that file access is restricted to the skill's root directory, preventing directory traversal attacks.
[PROMPT_INJECTION]: The script scripts/search_references.py contains a sanitization function sanitize_body that uses a regular expression to filter out common prompt injection patterns (e.g., 'ignore previous', 'system:', '') from reference content before it is displayed to the agent. This serves as a defensive measure against indirect prompt injection if reference files were to be modified by an untrusted party.
[COMMAND_EXECUTION]: The provided Python script uses standard libraries and does not invoke external shell commands or perform unsafe dynamic code execution.

webflow-enterprise-api