Deep Research

Content updated.

Content updated.

Content updated.

Display title updated.

Display title updated.

- Skill has been renamed from "research-cog" to "deep-research-cellcog" for improved clarity and alignment. - Updated DeepResearch Bench ranking reference from April 2026 to July 2026, and clarified that rankings are updated frequently. - Removed the outdated skill-card.md file. - Minor improvements to headings, section naming, and prompts for better consistency. - No changes to API usage or research capability—documentation and naming only.

- Added explicit requirements for Python 3 and the CELLCOG_API_KEY environment variable in SKILL metadata. - No code or logic changes; documentation/metadata update only.

- Expanded description to highlight support for financial analysis, crypto research, and news intelligence. - Clarified agent usage instructions: separated examples for OpenClaw and other agents. - Minor documentation improvements for accuracy and clarity.

- Documentation updated for clarity and completeness, especially on usage instructions and integrations. - Installation and example usage for Cursor and other agents improved, including code imports. - Skill description refined for conciseness and accuracy. - Minor formatting and organizational enhancements throughout the documentation.

**This update improves clarity, usage instructions, and research focus.** - Simplified and clarified skill description and instructions for faster onboarding. - Separated and streamlined SDK usage examples for OpenClaw and other agents. - Moved detailed SDK and setup references to the start, making first steps clearer. - Improved examples and tips for specifying research prompts and output formats. - Reduced redundancy and removed extraneous information for easier reading. - Retained all use case, feature, and prompt examples while making them more accessible.

- Major documentation update with detailed research use cases and example prompts. - Added sections covering competitive analysis, market research, investment analysis, academic research, and due diligence examples. - Explained output formats (interactive HTML, PDF, markdown, plain text) with recommendations. - Expanded guidance on chat modes: agent, agent team, and agent team max. - Provided advanced tips for citation handling, structuring analysis, and improving research quality. - Included leaderboard recognition: #1 on DeepResearch Bench (Apr 2026).

Version 1.0.9 - Major rewrite of documentation for clarity and brevity - Expanded description to highlight more research and output types - Added clear lists of supported research areas and output formats - Presented core multi-source and citation features more prominently - Clarified citation policy and chat mode usage - Added references to related skills for finance, crypto, data, and news analysis

- Added OpenClaw agent usage instructions and example to the SDK setup section. - Clarified differences between "OpenClaw agents" (fire-and-forget) and other agent invocation methods. - Updated SDK example code blocks for more precise usage guidance. - Minor editorial updates and formatting for improved clarity.

- Updated the Quick Pattern section to a simplified "Quick start" example, providing a basic usage pattern. - Added clear guidance to refer to the `cellcog` skill for full SDK/API details, including delivery modes and advanced usage. - Removed detailed asynchronous usage instructions to streamline documentation and avoid redundancy with the main CellCog skill. - Clarified separation of research-cog capabilities from the core SDK/API setup.

- Updated DeepResearch Bench achievement date to April 2026 in all mentions. - No functional or instructional changes; documentation remains consistent except for the updated benchmark recognition.

- Refined the skill description for clarity and added additional use cases and output formats. - Added a homepage link and specified supported operating systems in metadata. - Updated the description to emphasize multi-source synthesis and broader research capabilities (including due diligence and literature reviews). - No behavioral or functional changes to the skill code; documentation update only.

Expanded guidance on research chat modes - Added a detailed table describing when to use `"agent"`, `"agent team"`, and the new `"agent team max"` chat modes for different research scenarios. - Provided recommendations for scenarios requiring the highest accuracy, such as high-stakes due diligence or cutting-edge academic research. - Clarified that `"agent team"` remains the default for most research, but `"agent team max"` is available for institutional-grade analysis (requires ≥2,000 credits). - No code or interface changes—documentation update only.

- Added author information and explicit dependency listing for `cellcog` in the skill metadata. - Minor improvements to the prerequisites section, clarifying that `cellcog` is required for SDK setup. - No functionality changes; documentation and metadata only.

- Added "#1 on DeepResearch Bench (Feb 2026)" accolade to the description and introduction. - Included leaderboard link for transparency: https://huggingface.co/spaces/muset-ai/DeepResearch-Bench-Leaderboard. - No functionality or API changes; documentation update highlighting recent research benchmark performance.

- Added OpenClaw metadata with an emoji identifier. - Updated SDK usage instructions for v1.0+, introducing a new fire-and-forget research pattern with notification-based completion (no polling required). - Clarified that citations are not provided automatically and must be explicitly requested in the prompt, detailing how to format/position them. - Improved documentation for data accuracy, output formats, and example prompts. - Minor simplifications to setup instructions and SDK references.