competitor-monitoring-system

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests content from various untrusted third-party sources.
  • Ingestion points: The skill uses blog-scraper, linkedin-profile-post-scraper, twitter-scraper, reddit-scraper, and review-scraper to bring external content into the agent context.
  • Boundary markers: There are no boundary markers or instructions defined to prevent the agent from treating data found in scraped content as system instructions.
  • Capability inventory: The skill possesses file-writing capabilities used to generate baseline documents and intelligence reports in the clients/ directory.
  • Sanitization: No evidence of sanitization, filtering, or instruction-ignoring logic is present for the ingested external content.
  • [EXTERNAL_DOWNLOADS]: The skill is designed to systematically fetch data from external platforms including Meta Ad Library, Google Ad Library, G2, Capterra, and social networks.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 06:03 PM