The Agent Skills Directory

[SAFE]: The skill implements a structured 6-phase deliberation protocol with explicit boundaries between different agent personas to prevent groupthink and context bleed. No evidence of prompt injection or malicious instructions was found.
[SAFE]: Data management is handled through local memory files in the memory/ directory. The skill does not access sensitive system paths (e.g., SSH keys, credentials) or perform any network operations to non-whitelisted domains.
[SAFE]: All major decisions require explicit human approval in Phase 5 ('Human in the Loop'), providing a critical security checkpoint before any data is committed to the long-term decision log.

board-meeting