yao-geo-tracking
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the agent to execute a bundled Python script (scripts/render_yao_geo_tracking.py) to generate HTML and DOCX report files. The script uses standard Python libraries (argparse, json, zipfile, etc.) and implements sanitization to prevent injection in the generated documents.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content retrieved from external company websites to populate its tracking plans, creating a potential surface for indirect prompt injection. 1. Ingestion points: Official company websites and external business features (SKILL.md). 2. Boundary markers: The workflow defines an evidence-sourcing process (references/evidence-sourcing-workflow.md) and explicit official site validation. 3. Capability inventory: Subprocess execution of local scripts and file system writes for report generation. 4. Sanitization: The rendering script uses html.escape and xml.sax.saxutils.escape to sanitize output content.
Audit Metadata