illustrated-paper-guide
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and interpret external, untrusted academic papers and PDFs provided by the user. * Ingestion points: User-provided academic papers and PDF materials (SKILL.md). * Boundary markers: The instructions include clear directives to distinguish paper-supported evidence from agent interpretation and strictly forbid the invention of citations. * Capability inventory: The skill enables file system writing (Markdown guides and checklists), diagram generation via companion skills, and script execution for figure processing. * Sanitization: The skill relies on model-level controls and logical separation instructions rather than technical sanitization of document content.
- [DYNAMIC_EXECUTION]: The skill instructions involve the creation and execution of scripts for image extraction and figure cropping. * Evidence: SKILL.md and figure-reading-rubric.md describe storing and re-running scripts in a scripts/ folder to validate crop coordinates and ensure reproducible outputs.
Audit Metadata