xhs-competitor-research
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, which inherently introduces a surface for indirect prompt injection.
- Ingestion points: The skill reads user-provided CSV, XLSX, and JSON files via
scripts/analyze_attachment.pyand extracts content from public Xiaohongshu web pages using automated browser controls. - Boundary markers: Policy documents and instructions in
SKILL.mdrequire the agent to clearly state that the analysis is derived strictly from the provided samples and to avoid faking data or conclusions. - Capability inventory: The skill executes local Python scripts, performs browser-based scraping, and utilizes the
lark-clitool for exporting reports to Feishu. - Sanitization: Data is normalized and statistically audited by a Python script, and the agent is instructed to use structured frameworks (semantic coding) to process natural language content under governed research logic.
Audit Metadata