god-research-review
Pass
Audited by Gen Agent Trust Hub on Jun 20, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill mentions and provides templates for interacting with the Semantic Scholar API (
api.semanticscholar.org) to build citation graphs. This is a well-known service for academic research, and the usage is documented for legitimate literature traversal purposes. - [PROMPT_INJECTION]: While the skill uses terminology like 'Researcher-Warrior Mindset' and 'trained adversary', these are pedagogical metaphors for critical thinking and academic skepticism. They do not constitute an attempt to bypass agent safety guidelines or override system instructions.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external academic papers, which presents a potential surface for indirect prompt injection.
- Ingestion points: Reads research papers and technical documents provided by the user or fetched via search (SKILL.md Section 8 and 9).
- Boundary markers: The skill includes 'Anti-hallucination rules' and a 'Self-Review Checklist' to ensure the agent maintains factual accuracy and adheres to the review methodology.
- Capability inventory: Performs network requests to well-known academic APIs and search engines.
- Sanitization: The skill enforces a rigorous structure for reviews (Section 7 and 11) that acts as a logical filter for processed content.
Audit Metadata