Skills
Navigate
skills/azhi-ss/academic-figure-skills/academic-figure-architecture-extractor/Gen Agent Trust Hub

academic-figure-architecture-extractor

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The bundled script scripts/extract_pdf_figures.py executes system binaries pdfimages and pdftoppm via the subprocess module. While these calls use argument lists to mitigate shell injection, they represent a significant capability within the skill's environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from PDFs and paper URLs. The analysis reveals the following evidence chain: 1. Ingestion points: PDF files and URLs processed by the agent. 2. Boundary markers: The missing-info-policy.md and SKILL.md invention ban provide instructions to limit output to evidence. 3. Capabilities: Subprocess execution and local file writing. 4. Sanitization: No explicit filtering of text content from analyzed figures is present.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 03:34 AM
Security Audit — agent-trust-hub — academic-figure-architecture-extractor