ai-answer-trace
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to fetch sibling skill files from the vendor's GitHub repository (
https://raw.githubusercontent.com/XPOZpublic/xpoz-agent-skills/main/skills/<name>/SKILL.md). - [COMMAND_EXECUTION]: Provides and instructs the execution of several standalone Python scripts (
trace-claude.py,trace-openai.py,trace-gemini.py,cited-domains.py) to interface with AI APIs and process results. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted web grounding data returned by third-party AI engines.
- Ingestion points: The
trace-*.pyscripts ingest search results, including page titles and URLs, from the Claude, OpenAI, and Gemini APIs. - Boundary markers: None identified in the script logic for processing search results.
- Capability inventory: The provided scripts execute network requests to AI service providers and resolve redirects for web URLs using standard libraries.
- Sanitization: The scripts perform data reformatting and JSON serialization without explicit sanitization of the content retrieved from search results.
Audit Metadata