deep-research-pro

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

• Third-party content exposure detected (high risk: 0.90). SKILL.md Step 3/Step 4 explicitly instruct the agent to run a DuckDuckGo search script (/home/.../ddg) and fetch full pages with curl for "deep-read" of arbitrary web/news URLs, meaning the agent ingests untrusted, potentially user-generated third‑party content that can directly influence its synthesis and next actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches arbitrary external webpages at runtime via the command pattern curl -sL "" | python3 -c "..." which pipes remote content into the agent's processing (i.e., the fetched text is ingested and can directly influence prompts/instructions), so this runtime fetch of external URLs is a high-confidence risk.