research

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

• Third-party content exposure detected (high risk: 0.80). The skill's SKILL.md Interactive Research workflow explicitly instructs the agent to "Do the research
• Web search, fetch docs, explore code" (Research Loop step 1), so the agent will fetch and read open/public third‑party content (webpages/docs) and act on those findings to update docs and choose next actions, exposing it to indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's parallel-research CLI makes runtime calls to the Parallel AI API at https://api.parallel.ai/v1 and consumes the returned markdown/result as the research output (used to populate docs and drive the agent's responses), so external content from that URL directly controls agent output and is a required dependency for deep research.