agent-reach

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

• Third-party content exposure detected (medium risk: 0.30). The workflow monitors and consumes public social feeds, comments, and issue trackers such as Reddit, Twitter/X, and GitHub issues where outsiders can post arbitrary text without prior selection.

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

• Potentially malicious external URL detected (medium risk: 0.30). The skill instructs the agent to fetch a remote installation guide from a public GitHub repository URL during runtime setup, which constitutes an unverifiable unofficial operational dependency.