The Agent Skills Directory

[SAFE]: The skill provides a methodology for multi-agent workflows focusing on security best practices like least privilege, isolation, and explicit approval for consequential actions.
[COMMAND_EXECUTION]: The skill utilizes a bundled script, scripts/validate_plan.py, for structural verification of orchestration plans. A review of the script confirms it uses only standard Python libraries to perform schema validation and dependency checks, with no evidence of malicious behavior, network access, or unauthorized file system operations.
[DATA_EXFILTRATION]: The instructions promote secure handling of credentials, explicitly directing agents not to broadcast secrets in handoffs or shared states and to use isolated environments for sensitive context.

multi-agent-orchestration