The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests data from external sources, creating a surface for indirect prompt injection. However, this is inherent to its documented purpose and mitigated by its testing strategy.
Ingestion points: Content is pulled from shared document platforms (Google Drive, SharePoint) and messaging channels (Slack, Teams) as specified in the 'Stage 1: Context Gathering' section of SKILL.md.
Boundary markers: The instructions do not specify technical delimiters for ingested data, but they do implement a functional boundary by testing the final document in a fresh environment.
Capability inventory: The skill uses create_file and str_replace to manage documentation artifacts based on the ingested content.
Sanitization: External content is used to inform the drafting process without explicit sanitization steps mentioned.
[SAFE]: The skill implements security best practices by recommending a 'Reader Testing' phase in Stage 3, which uses an isolated sub-agent (a fresh Claude instance with no conversation context) to review the generated document. This process effectively identifies blind spots, context bleed, and potential contradictions before finalization. All external service references (Slack, Teams, Google Drive, SharePoint) are well-known services.

doc-coauthoring