competitor-analysis

Pass

Audited by Gen Agent Trust Hub on Jul 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface, which is inherent to its core functionality of processing external data.
  • Ingestion points: External content from competitor websites, review platforms (G2, Capterra, Trustpilot), and ad transparency libraries is ingested via WebFetch and WebSearch (SKILL.md).
  • Boundary markers: The skill does not currently specify the use of delimiters or specific instructions to ignore embedded prompts within the retrieved data.
  • Capability inventory: The toolset is limited to search and retrieval functions (WebSearch, WebFetch); it does not include dangerous capabilities such as arbitrary code execution or filesystem write access.
  • Sanitization: There is no mention of sanitization or escaping of the interpolated web content.
  • [DATA_EXFILTRATION]: The skill retrieves information from well-known services and public websites (e.g., LinkedIn, Crunchbase, Meta Ad Library). This is documented behavior for competitive research and does not target sensitive user environment variables or private credentials.
  • [EXTERNAL_DOWNLOADS]: The skill performs runtime retrieval of content from external URLs using WebFetch to analyze pricing, product features, and messaging of target competitors.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 12, 2026, 12:36 AM
Security Audit — agent-trust-hub — competitor-analysis