The Agent Skills Directory

[SAFE]: The skill establishes a governance model for AI agent operations, emphasizing the 'Separation of Duties' principle to ensure that generating agents cannot bypass their own verification standards.
[COMMAND_EXECUTION]: The document references internal utility commands for auditing and state transitions (e.g., lp task move, agent_switch.py audit). These are used for process tracking and do not perform unsafe or arbitrary operations.
[INDIRECT_PROMPT_INJECTION]: The 'Independent Adversarial Review' process is designed as a security control to identify logic errors and potential exploits in agent-generated content before it enters production.

acceptance-gates