improve-react

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

• Third-party content exposure detected (high risk: 0.75). Yes: the workflow requires fetching React Doctor “canonical fix recipes” from public web URLs (e.g., https://www.react.doctor/prompts/rules/<plugin>/<rule>.md) and inlining them into the agent’s context while writing plans, which is outsider-authored free text ingested at runtime via the LLM prompt.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly runs remote code at runtime via npx react-doctor@latest --json ... and requires fetching/inlining canonical prompts from https://www.react.doctor/prompts/rules//.md, which both execute remote code or directly control the agent's prompts.