tzai-image

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs using and asking for API keys (examples show --api-key 'sk-YOUR_TOKEN' and "copy-paste init" / one-shot --api-key on generate), which requires the agent to accept and then embed secret values verbatim in commands — an insecure exfiltration pattern.

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/tzai-image → cmd_generate, the engine ingests outsider-authored free text via the --prompt argument (bound from user/slash input) and sends it directly to TaoziAPI in the POST body as prompt (no “select specific item to fetch” step).