doc-coauthoring

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection due to its core functionality of reading and processing data from external, potentially untrusted sources.
  • Ingestion points: The skill explicitly reads content from Slack, Teams, Google Drive, SharePoint, and shared document links provided by the user (as described in Stage 1).
  • Boundary markers: The agent is not instructed to use unique delimiters (like XML tags or specific block markers) or safety pre-prompts when ingesting external content to prevent the agent from following instructions embedded within that data.
  • Capability inventory: The skill is capable of modifying the local filesystem using tools like create_file and str_replace, and it can invoke sub-agents for automated testing, which could be exploited if malicious instructions are ingested.
  • Sanitization: The instructions do not define any validation, filtering, or escaping logic for content retrieved from external integrations.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 04:00 PM
Security Audit — agent-trust-hub — doc-coauthoring