bannerbear

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

• Third-party content exposure detected (high risk: 0.70). The SKILL.md explicitly lists a "Create Screenshot — Capture a screenshot of a public webpage" action, which makes the agent fetch and ingest untrusted public web content (third-party webpages) as part of its runtime workflow and that output can be used to drive subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.90). This skill requires installing and running the Membrane CLI via "npm install -g @membranehq/cli@latest" (which pulls and executes code from the npm registry, e.g. https://registry.npmjs.org/@membranehq/cli), a runtime external dependency that fetches and executes remote code and is required for the skill to operate.