The Agent Skills Directory

[SAFE]: The skill provides structured instructions for task decomposition and multi-tool coordination without any executable code or dangerous network operations.
[SAFE]: Includes explicit safety guidelines (Red/Yellow/Green categories) that require user confirmation for high-risk or large-scale actions such as irreversible deletions or production data access.
[SAFE]: Implements structured error handling, retry logic, and conflict resolution strategies to ensure process integrity and user oversight.
[SAFE]: While the orchestration of external data processing (e.g., web scraping) creates a surface for indirect prompt injection, the skill's requirement for user approval of the decomposition plan and the inclusion of manual confirmation for high-risk steps serve as effective architectural mitigations.

mcp-conductor