Deep Research Pro

**Judgement:** Deep Research Pro is a strong, well-structured deep-research skill that fits OpenClaw-style agent workflows and is well-suited as both a primary research tool and a spawned sub-agent for complex tasks. **Strengths:** - Implements a full research loop: clarify → plan sub-questions → multi-query web+news search → deep-read key pages → synthesize. - Enforces citation for every claim and explicitly flags single-source or low-confidence findings. - Produces highly usable outputs with a consistent report template, executive summary, key themes, key takeaways, and methodology. - Self-contained: relies on DuckDuckGo search scripts, `curl`, and `python3`, avoiding external paid APIs. **Key Risks / Limitations:** - Depends heavily on external web search and site availability; failures or blocking at the network/search level will degrade results and there is no explicit error-handling or retry logic. - No explicit content-safety filtering beyond general epistemic caution; topics that surface harmful or sensitive material rely on upstream platform safeguards. - Performance can be slower on broad topics due to multiple searches and deep-reading several pages; may be overkill for very simple queries. - Hardcoded paths and a specific model name in the sub-agent example may need adjustment in some deployments. **Recommended Scenarios:** - Best for users who need thorough, source-backed briefs: market/industry overviews, technology and policy analyses, current-events deep dives, and research to support decision-making or writing. - Particularly useful as a delegated sub-agent in larger workflows that require a dedicated research phase and a saved markdown report. - Less ideal for ultra-fast, lightweight lookups where a single quick answer is sufficient, or for domains requiring specialized proprietary databases beyond the open web.