research
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from external websites (official documentation, source code, and specifications). This ingestion point represents an attack surface where malicious instructions could be embedded in the technical data. The skill mitigates this by explicitly instructing the agent to treat fetched source content as evidence rather than instructions (Anti-pattern #27).
- Ingestion points: Web content from Tier 1 and Tier 2 authoritative sources as described in SKILL.md and RESEARCH-FORMAT.md.
- Boundary markers: The skill contains explicit instructions to distinguish between gathered evidence and actionable instructions.
- Capability inventory: The skill has permissions to read project dependency manifests (package.json, go.mod, Cargo.toml, requirements.txt, etc.) and write markdown files to the docs/research/ directory.
- Sanitization: The agent is guided to use external content only for technical comparison and verdict generation within the defined markdown template.
Audit Metadata