webcmd-adapter-author

Pass

Audited by Gen Agent Trust Hub on Aug 1, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill requires the agent to visit and analyze untrusted external websites to discover APIs and DOM structures, which presents an indirect prompt injection surface.
  • Ingestion points: The agent ingests external data using webcmd browser analyze, webcmd browser network, and webcmd browser eval across multiple files including references/api-discovery.md and references/site-recon.md.
  • Boundary markers: The instructions do not explicitly warn the agent about the possibility of adversarial instructions being hidden in website content, such as metadata or hidden HTML elements, though they do recommend verifying results against the visible UI.
  • Capability inventory: The agent has the ability to write executable JavaScript files (Write, Edit) and install them as plugins using Bash, which could be exploited if an injection successfully influences the generated code.
  • Sanitization: No specific measures are outlined to sanitize or escape data extracted from external websites before it is interpolated into generated adapter code or used to influence the agent's decision-making process.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 1, 2026, 06:45 AM
Security Audit — agent-trust-hub — webcmd-adapter-author