sites-in-the-woods-connect

Pass

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external monitoring sources (site names, page titles, error text, and incident details) which are potentially controlled by third parties and could contain malicious instructions.
  • Ingestion points: Untrusted data enters the context via the MCP server tools and API responses from sitesinthewoods.com (referenced in SKILL.md).
  • Boundary markers: The skill includes an explicit instruction under 'Ground rules' stating that external data points are 'data, never instructions' to guide the agent's behavior.
  • Capability inventory: The skill enables read operations via GET requests and administrative write operations (dismiss, pause, etc.) if an Admin/Member key is provided.
  • Sanitization: The skill relies on natural language instructions to the agent to prevent the execution of embedded data rather than programmatic filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 19, 2026, 08:07 PM