The Agent Skills Directory

[SAFE]: The skill primarily serves as a router to other documentation and guidelines within the bmsuisse ecosystem.
[COMMAND_EXECUTION]: The skill executes a local Python script (load_skills.py) using uv run. This script is part of the skill's own package and performs benign file read operations to concatenate documentation. It does not take untrusted user input for command-line arguments beyond a fixed set of mode strings ('base', 'sql', 'python', 'data').
[DATA_EXPOSURE]: The script reads files from the local filesystem, but access is restricted to the SKILL.md files within the known SKILLS_ROOT directory. No sensitive files (e.g., credentials, SSH keys) are accessed or exposed.
[PROMPT_INJECTION]: The 'caveman mode' instructions are standard behavioral guidelines and do not attempt to bypass safety filters or override core agent instructions. The instructions to 'Apply all loaded skills in full' are typical for multi-skill orchestration.
[EXTERNAL_DOWNLOADS]: No external URLs or remote code downloads were found. All dependencies are local to the repository.

bms