interhuman-stream

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill requires an API access token as an explicit input and instructs embedding that token verbatim in connection headers or the Sec-WebSocket-Protocol (shown in code examples), so an agent would need to handle and could be asked to output the secret value directly.

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

• Third-party content exposure detected (high risk: 0.90). The skill connects to the public WebSocket endpoint wss://api.interhuman.ai/v0/stream/analyze and is required to relay all WebSocket messages exactly as received, meaning it ingests and outputs untrusted third-party JSON produced by the external Interhuman service (potentially derived from user-provided video) without filtering or modification.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill opens a runtime WebSocket to wss://api.interhuman.ai/v0/stream/analyze and mandates relaying all server messages verbatim, so remote content received at that URL can directly control agent outputs or inject instructions.