doc-coauthoring

Pass

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references the use of Model Context Protocol (MCP) servers and platform-native connectors to access data from well-known and trusted services such as Slack, Google Drive, and SharePoint.
  • [PROMPT_INJECTION]: The skill processes content from external documents and user-provided context (info dumps), presenting a surface for indirect prompt injection common to documentation tasks. This is handled as an inherent risk of the document-writing use case.
  • Ingestion points: Shared documents and team channels mentioned in initial-questions.md and info-dumping.md.
  • Boundary markers: None identified in the instructional flow.
  • Capability inventory: File modification via create_file and str_replace in step-5-drafting.md.
  • Sanitization: Content is processed directly through the agent conversation without explicit sanitization steps.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 8, 2026, 02:10 PM
Security Audit — agent-trust-hub — doc-coauthoring