academic-paper-writing
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external manuscripts, academic reviews, and bibliographies, creating a surface for indirect prompt injection where instructions embedded in these documents could attempt to influence the host agent's behavior.
- Ingestion points: Manuscript files (PDF, TeX, MD, TXT), experiment records (CSV, JSON), and bibliography files (BibTeX, JSON) loaded through materials.py and prepared for the agent.
- Boundary markers: The skill includes explicit instructions in tasks.py and workflow.py informing the host agent to treat all supplied files as data rather than instructions and to disregard any embedded commands.
- Capability inventory: The skill's scripts are restricted to local file processing and structured data validation. It lacks network capabilities and does not perform dynamic code execution on external data.
- Sanitization: Data is validated against a JSON Schema in workflow.py and processed using specific parsers for BibTeX and CSV content in materials.py.
Audit Metadata