cli-gpt-image

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime path takes the Codex backend’s SSE response events (including response.output_item.done with item.revised_prompt and other text fields) and returns them in the agent-visible JSON envelope on stdout, so outsider-authored free text from ChatGPT/Codex can enter the LLM context via the tool output.