The Agent Skills Directory

[COMMAND_EXECUTION]: The skill incorporates a local Python script (scripts/validate_round.py) which the agent is instructed to execute using python3 to verify that session manifests and file structures comply with the isolation protocol.
[COMMAND_EXECUTION]: The unit tests provided in tests/test_validate_round.py use the subprocess module to invoke the validation script for automated verification.
[DATA_EXPOSURE]: While the skill manages and reads various local files within a dedicated session directory, its 'manifest' system and validation logic are designed to enforce a white-list approach to file access, preventing unauthorized data leakage between independent sub-agents.

cuihuo-brainstorm