power-research
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves fetching arbitrary content from the web and storing it in a local directory for analysis. This content is then processed to extract claims and identify research gaps, creating a surface where malicious instructions embedded in external web pages could attempt to influence the agent's research process or final synthesis.
- Ingestion points: Web content is fetched and saved to markdown files in the sources/ directory as described in the 'COLLECT Mode' section of SKILL.md.
- Boundary markers: The skill utilizes YAML frontmatter for source metadata, but it lacks explicit boundary markers or instructions to ignore embedded prompts within the main body of the fetched source text.
- Capability inventory: The agent possesses capabilities for web searching, content fetching, local filesystem writes, and the execution of internal utility scripts (scripts/init.py and scripts/index.py).
- Sanitization: There is no evidence of sanitization or filtering of fetched web content before it is read by the agent to perform claim extraction.
Audit Metadata