adversarial-python

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

• Third-party content exposure detected (high risk: 0.85). The required workflow fetches public web content at runtime (“fetch https://docs.python.org/3/whatsnew/3.{N}.html” for version-delta briefing), and that fetched free text is inserted into the reviewer prompt that is sent to the LLM.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly instructs at runtime to fetch the What's New pages and inject their content into the reviewer prompt (e.g. fetch https://docs.python.org/3/whatsnew/3.{N}.html to compose the delta briefing), so remote documentation controls the composed reviewer prompt.