imagine

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

• Third-party content exposure detected (medium risk: 0.30). Source: user-supplied --prompt is taken from CLI args (outsider free text) and is directly embedded into the runtime request body (input_text / input) sent to the local OAuth proxy in scripts/generate.js and scripts/edit.js.

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

• Potentially malicious external URL detected (high risk: 0.80). The scripts spawn and require the npx-installed OAuth proxy at runtime (e.g. "npx openai-oauth --port 10531") and direct users to run "npx @openai/codex login", both of which fetch and execute remote npm package code that the skill depends on at runtime.