document-metadata-review
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill follows security best practices by explicitly prohibiting the upload of sensitive documents to external services for processing. It also restricts the use of specific external services and git operations.
- [PROMPT_INJECTION]: The skill processes untrusted document formats (PDF, DOCX, PPTX, XLSX) which could contain malicious instructions (indirect prompt injection). The instructions mitigate this by focusing on inspection and flagging rather than execution, and include anti-patterns specifically advising against treating external content as trusted instructions.
- Ingestion points: Document files (PDF, DOCX, PPTX, XLSX) processed through local tools.
- Boundary markers: No specific delimiters or warnings for the agent to ignore document content are present in the core instructions.
- Capability inventory: Access to local documents, shell tools, and editor tools.
- Sanitization: No explicit sanitization or validation of the document content is described.
Audit Metadata