The Agent Skills Directory

[SAFE]: The skill defines a logical workflow for multi-perspective analysis using sub-agents. It does not contain any executable scripts, binary files, or external network requests.- [SAFE]: Persistence guidelines within the skill explicitly discourage writing to ad-hoc or shadow paths like ~/.claude/notes, instead recommending the use of official tools or session memory, which reduces the risk of unintended data leakage or unauthorized file modifications.- [SAFE]: The prompt templates provided for sub-agents are well-structured and focused on specific decision criteria. They do not attempt to bypass system safety guidelines or override agent constraints.- [SAFE]: The skill's data ingestion process is limited to gathering relevant codebase context for manual analysis. It lacks any automated execution or write capabilities that could be exploited through indirect prompt injection.

council