speech-to-text-with-speakers

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

• Third-party content exposure detected (high risk: 0.75). Outsider-authored free text can enter the LLM context via the runtime transcription result (transcribed speech from an uploaded/public_url audio source, i.e., not authored by the operating user), which is returned as inline text/vtt/json content.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly requires fetching live schema/instructions at runtime from https://api.agentpmt.com/mcp/ (and invoking the REST endpoint https://api.agentpmt.com/products/purchase), and those returned instructions/schemas are treated as authoritative and directly control the tool's behavior, so this is a runtime external dependency that can affect prompts.