second-opinion-review
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user drafts, critiques, and external web content fetched via the fetch tool (SKILL.md Step 1, fact-check-short.md item 2). It mitigates injection risks by including explicit instructions to treat all external content as data and to ignore any embedded instructions. This satisfies the requirement for boundary markers and instruction sanitization. The capabilities available to the agent for this task include script execution and network fetching, which are appropriate for the skill's purpose.
- [DYNAMIC_EXECUTION]: The skill uses dynamic Python execution for word counting (python3 -c) and executes a local script (scripts/claims_scan.py) for text analysis (SKILL.md Step 3, Step 8). These executions are strictly scoped to the skill's primary purpose of draft review and analysis, representing a standard use of helper scripts and basic command-line utilities.
Audit Metadata