compdf-pdf-to-json
Pass
Audited by Gen Agent Trust Hub on Aug 15, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes PDF documents, creating an attack surface where untrusted data enters the agent context. Ingestion points: Source PDF files (SKILL.md). Boundary markers: Absent for input content; results are structured as JSON. Capability inventory: Network requests to ComPDF API; file system reads/writes for local API keys. Sanitization: None specified for external document content.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill manages an API key in a local file named
api_key. This is documented as a safe practice for local state management, and the skill includes explicit instructions to avoid displaying or leaking this key in outputs or logs. - [SAFE]: The skill mandates explicit user confirmation before any document uploads or operations involving file deletion or encryption, providing a necessary layer of human oversight.
Audit Metadata