xy-roundtable
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a bundled Python script (
scripts/atoms-search.py) to search and retrieve data from a local knowledge base for use in the expert simulations. - [EXTERNAL_DOWNLOADS]: The
atoms-search.pyscript makes network requests toapi.xyskill.xyz. This is a vendor-owned endpoint used to query the full atomic knowledge base. The behavior is documented as a normal part of the skill's search functionality. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (business topics) and interpolates them into expert personas and reports.
- Ingestion points: User-provided business topics (议题) in the roundtable and deep analysis modes.
- Boundary markers: The skill includes instructions to first "press" (压题) the topic into a standardized single-line format before proceeding with analysis.
- Capability inventory: The skill has the ability to write files to the local directory (report generation) and execute a bundled Python script via subprocess.
- Sanitization: The Python script uses standard argument parsing to handle keywords passed from the agent context.
Audit Metadata