review-reputation

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection Surface: The skill is designed to ingest and analyze untrusted data from external sources like public reviews, emails, and support tickets. This creates a surface where malicious instructions could be embedded in the content to influence the agent's behavior.
  • Ingestion points: Data is fetched from public web pages (Google, Yelp) and various CRM, payment, and email connectors as described in SKILL.md and reference/sources.md.
  • Boundary markers: The skill includes a 'What not to do' section in SKILL.md that explicitly instructs the agent to treat ingested text as data rather than commands.
  • Capability inventory: The agent uses WebFetch for data gathering and generates drafts for public review responses and customer win-back offers.
  • Sanitization: The skill relies on mandatory human-in-the-loop approval gates for all outbound actions, which serves as a primary control against malicious output generated via injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:06 PM
Security Audit — agent-trust-hub — review-reputation