consult-governance
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Google Drive folder names, document titles, and CRM deal notes during its audit process.
- Ingestion points: Data is retrieved from Google Drive via
mcp__google-workspace__list_drive_items(SKILL.md Step 2) and CRM tables viamcp__zynkr__list_dealsandmcp__supabase__execute_sql(SKILL.md Step 3). - Boundary markers: The instructions do not define clear delimiters or "ignore" instructions for processed external strings, potentially allowing malicious content in folder names or notes to influence agent behavior.
- Capability inventory: The skill utilizes read-only tools for Google Workspace and Supabase; no write access is requested or utilized, limiting the impact of any successful injection.
- Sanitization: Folder names are validated against a strict regex pattern
^\[(\d+)\] (.+?)（(.+)）$, which provides structural validation and mitigates basic exploitation attempts.
Audit Metadata