superagent

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

• Third-party content exposure detected (low risk: 0.10). It only ingests outsider-authored free text when the user explicitly selects specific outsider-originated targets (e.g., “scan/check this URL, file, email, skill, or MCP repository” via the corresponding Context Guardrails /api/v1/context/* endpoints) rather than via any always-on queue/feed the workflow monitors.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs MCP clients to connect at runtime to https://www.superagent.sh/mcp and to call the Superagent Context Guardrails endpoints so that the remote server's responses (tool lists and API verdicts) directly control agent enforcement/behavior, making this an external runtime dependency that can affect prompts or execution.