academic-figure-analyzer
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as research PDFs, code repositories, and paper drafts, which serves as a potential vector for indirect prompt injection.
- Ingestion points: Data enters the agent's context through PDF files (processed via an included script), repository files, and external URLs.
- Boundary markers: The skill relies on specific output schemas and a human-readable strategy defined in reference files.
- Capability inventory: The agent can run a local Python script that calls system utilities and uses common libraries to manipulate files.
- Sanitization: The instructions in 'references/missing-info-policy.md' explicitly mandate that the agent use placeholders for uncertain information and avoid inventing architectural details not present in the source material.
- [COMMAND_EXECUTION]: The skill includes a script, 'scripts/extract_pdf_figures.py', which executes system commands for PDF processing.
- Evidence: The script uses 'subprocess.run()' to invoke utilities like 'pdfimages' and 'pdftoppm'.
- Implementation Details: The execution is implemented safely by passing arguments as a list to avoid shell interpolation. The script also performs validation on its inputs, such as page ranges and DPI settings, and uses path resolution for file locations.
Audit Metadata