zendesk-guide

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill's SKILL.md explicitly instructs using Membrane to proxy requests to the Zendesk Guide API (e.g., "membrane request CONNECTION_ID /path/to/endpoint") and to run actions that fetch Guide content (articles/comments), which are third‑party, user‑generated/untrusted content that the agent would read and could influence subsequent actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs running npx @membranehq/cli@latest (which fetches and executes the @membranehq/cli package from the npm registry, e.g. https://registry.npmjs.org/@membranehq/cli) at runtime and relies on that fetched code as a required CLI — meaning remote code is executed during runtime.