The Agent Skills Directory

[SAFE]: The skill implements a controlled three-stage workflow (Context Gathering, Refinement, and Reader Testing) for document authoring that adheres to standard operational procedures.
[SAFE]: Data ingestion is handled through authorized workspace integrations (e.g., Slack, Google Drive, Teams) and user-provided files, relying on the host environment's existing security permissions and connector settings.
[SAFE]: The 'Reader Testing' stage acts as a security and quality control mechanism by invoking isolated sub-agents in fresh instances. This process specifically checks for ambiguities, false assumptions, and contradictions, effectively mitigating the risk of accidental context bleed or instruction persistence from processed external data.

docs-coauthor