Parallel Deep Research

**Quick judgment**: Parallel Deep Research is a strong, specialized skill for comprehensive, citation-backed research tasks where depth and synthesis across many sources matter more than speed. It is best used when the user explicitly requests a thorough investigation or full report, not for everyday fact lookups or real-time news. **Recommended scenarios** - Complex research questions requiring synthesis from many sources (10+), such as market landscapes, regulatory overviews, or competitive analysis. - Due diligence and strategic decision support, where having structured findings plus traceable sources is essential. - Situations where you can afford longer runtimes (minutes to an hour) in exchange for depth and breadth of coverage. **Key risks & limitations** - **Speed vs. depth trade-off**: Deep tiers (ultra, ultra2x/4x/8x-fast) can take many minutes or longer; this is unsuitable for users expecting near-instant responses. - **Not ideal for breaking news**: The skill itself warns that current news or very recent events are better handled via `parallel-search` with date filters. Using this skill for fresh news may yield stale or incomplete information. - **External content quality**: Although reports include citations, the underlying web sources may contain inaccuracies or bias; downstream agents must not treat all sourced claims as ground truth without critical evaluation. - **API dependency**: Requires a Parallel API key and stable network; failures or rate limits from the external API will directly impact reliability. Overall, this skill fits well as the "heavy-duty research" backbone in an OpenClaw setup: call it when a user explicitly asks for deep, comprehensive research, then use other skills (or core reasoning) to interpret, summarize, or act on the returned report.