research-content

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill uses allowed tools (WebSearch, WebFetch, Read, Grep, Glob) strictly for their intended purpose of researching topics and analyzing public repositories.
  • [SAFE]: The provided reference files (sector-profiles.md and twitter-algorithm.md) contain purely informational text and do not include executable code or hidden instructions.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted data from the web and external GitHub repositories.
  • Ingestion points: Data enters via WebSearch and WebFetch (SKILL.md, Step 2) and repo analysis via Read/Grep/Glob.
  • Boundary markers: None identified; external content is processed as part of the research flow.
  • Capability inventory: The skill is restricted to information gathering and brief generation. It lacks the ability to execute commands (exec/eval), write to the file system (other than standard output), or perform sensitive network operations.
  • Sanitization: No explicit sanitization or filtering of external content is described, but the risk is mitigated by the lack of dangerous execution tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 10:05 AM