The Agent Skills Directory

[PROMPT_INJECTION]: The skill identifies internal traffic filters by ingesting extra_data from external tracking events. This represents an indirect prompt injection surface, though it is well-protected by validation logic.- Ingestion points: The extra_data field within the DuckDB event lake, processed in scripts/detect_employee_filter.py.- Boundary markers: Instructions in SKILL.md require the agent to output the script's one-line results verbatim, limiting the agent's interaction with the raw data values.- Capability inventory: The skill can read and modify the bucket.json file in the user's configuration directory.- Sanitization: The script uses a case-insensitive whitelist of specific keys (e.g., 'internal', 'employee') and strict boolean validation for values, preventing arbitrary content from being processed as a filter.

identify-employees