parallel-web-extract

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

• Third-party content exposure detected (high risk: 1.00). The skill's SKILL.md explicitly instructs the agent to run parallel-cli extract on arbitrary URLs ("Use for fetching any URL
• webpages, articles, PDFs...") and to return and act on the extracted verbatim content, which clearly ingests untrusted public web content (from $ARGUMENTS) into the agent's workflow.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls parallel-cli to fetch arbitrary external URLs supplied as $ARGUMENTS (e.g., "https://docs.parallel.ai" in the examples) and returns the fetched content verbatim into the agent response, meaning remote content can directly control prompts at runtime.