competitive-landscape-zhcn
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze data from untrusted external sources, creating a potential surface for malicious instructions to influence the agent.
- Ingestion points: Technical reports, patent data, research papers, and web search results retrieved via tools like Exa, Tavily, and Brave (documented in references/source-routing.md and SKILL.md).
- Boundary markers: There are no specific instructions for the agent to wrap external content in delimiters or use protective framing when processing results.
- Capability inventory: The agent is required to read and write local files (request.md, query_log.csv, report.md, etc.) to track research and generate deliverables.
- Sanitization: The instructions do not define methods for sanitizing or validating external content before it is interpolated into the agent's context.
- [COMMAND_EXECUTION]: The skill suggests the use of command-line tools for report generation.
- Evidence: references/deliverables.md suggests using pandoc, python-docx, and reportlab for converting Markdown to DOCX or PDF formats. While these are suggested as optional host-specific tools, they represent a surface for shell command execution.
Audit Metadata