The Agent Skills Directory

[SAFE]: The skill performs local directory scans to identify project structure and technology stacks. The accompanying script, dev_core.py, uses the Python standard library to iterate through file paths and extensions without reading file content or performing network communication.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted local file names as input. Ingestion points: dev_core.py (via pathlib). Boundary markers: Present in scan_text (headers and section labels). Capability inventory: None (no file-write, network, or subprocess calls). Sanitization: None (passive metadata reporting only). This implementation is safe as it does not execute the data or provide a path for escalation.

dev