qonto-receipt-hunter
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from untrusted external sources, creating a potential attack surface where malicious instructions embedded in emails or documents could influence agent behavior.
- Ingestion points: External data enters the context through Gmail message retrieval, Google Drive file searches, and Qonto transaction metadata (SKILL.md).
- Boundary markers: The workflow strictly mandates human-in-the-loop validation, requiring the agent to "Show the match (transaction ↔ document) and wait for explicit confirmation" for every individual item (SKILL.md).
- Capability inventory: The skill possesses write capabilities to the user's financial account via the
qonto.upload_attachmentandqonto.request_attachment_uploadtools (SKILL.md). - Sanitization: Risk is mitigated by explicit instructions to never process attachments in bulk and to require individual user approval for each document match, preventing automated execution of potentially malicious content found in retrieved files.
Audit Metadata