brand-research-agent
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted content from external websites to perform brand analysis.\n- Ingestion points: Website HTML and text content are fetched by the
scripts/analyze_brand.pyscript and passed to the specialized agents in theagents/directory.\n- Boundary markers: The skill provides structured JSON templates for agent outputs, which helps maintain format but does not fully prevent prompt-based manipulation of the analysis content.\n- Capability inventory: The skill can perform network requests to any URL and write the analysis results to a local file (brand_profile.json).\n- Sanitization: TheBrandDataExtractorin the Python script parses specific tags (headings, meta, CSS), providing a layer of filtering before the content reaches the agents.
Audit Metadata