An autonomous research copilot that manages the full scientific lifecycle from brainstorming to final reporting with a focus on reproducibility.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install meta-research
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install meta-research using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Meta-Research?
Meta-Research is a sophisticated autonomous agent designed to guide developers and researchers through a rigorous, error-correcting scientific pipeline. Built as one of the premier Openclaw Skills, it handles the complexities of brainstorming, literature reviews, experiment design, and data analysis. The skill emphasizes an audit-ready approach, ensuring every decision is documented with clear rationale to minimize bias and maximize transparency.
By operating with a non-linear state machine, Meta-Research allows for dynamic backtracking and pivoting when new evidence or novelty gaps are identified. This makes it an indispensable tool for anyone looking to professionalize their AI research or scientific exploration within a coding environment, providing a structured framework that reduces undocumented decisions and ensures technical truth at every stage.
Meta-Research Use Cases
- Brainstorming and scoring novel AI research directions using the FINER rubric.
- Conducting automated, systematic literature reviews to identify novelty gaps.
- Designing rigorous experimental protocols with built-in controls and metric locking.
- Executing data analysis pipelines with automated ablation studies and error analysis.
- Generating comprehensive research drafts that adhere to strict reproducibility checklists.
How Meta-Research Works
- Entry Point Identification: Upon invocation, the agent determines your current research stage and initializes a project LOGBOX.md to track the decision provenance trail.
- Exploration Management: It organizes the file system into exploration-specific directories, ensuring that failures and pivots are archived while preserving a clean workspace.
- Phase Execution: The agent cycles through five core phases (Brainstorming, Lit Review, Experiment Design, Analysis, Writing) using specialized instructions and templates.
- Dynamic Backtracking: Based on specific triggers—such as finding that an idea has already been solved—the agent intelligently transitions back to earlier phases while carrying forward reusable artifacts.
- Audit Logging: Every milestone, transition, and pivot is logged in a central registry, providing a complete history of the research trajectory for future review.
Meta-Research Setup
To integrate this skill into your environment, symlink the meta-research directory to your local skills path. This allows you to access one of the most powerful Openclaw Skills across your projects.
# Install as a personal skill
ln -s /path/to/meta-research ~/.claude/skills/meta-research
# Install as a project-specific skill
ln -s /path/to/meta-research /your/project/.claude/skills/meta-research
Invoke the agent directly using the command:
/meta-research [your research question or topic]
Meta-Research Data Schema & Taxonomy
Meta-Research maintains a structured taxonomy to ensure reproducibility and clarity across the research lifecycle:
| Component | Path | Description |
|---|---|---|
| Logbox | LOGBOX.md |
The central audit trail containing decision logs and the exploration registry. |
| Shared Resources | shared/data/ |
Immutable datasets and resource maps shared across multiple research paths. |
| Exploration Dir | explorations/NNN-slug/ |
Isolated environment for a specific research direction with its own code and artifacts. |
| Phase Artifacts | explorations/NNN/*.md |
Individual markdown files for each stage (e.g., brainstorm.md, protocol.md, analysis.md). |
Meta-Research Advanced Features
- Non-linear research state machine that handles pivots and forks without losing historical context.
- Active bias mitigation tools, including pre-registration of metrics and dataset splits before execution.
- Integrated 12-framework ideation engine for generating high-impact research questions.
- Automated PRISMA trail generation for transparent literature screening and evidence mapping.
- High-autonomy mode for paper searching and template drafting with user-controlled checkpoints at critical transitions.
SKILL.md
Loading
Related Openclaw Skills
Huangdi Health Timer
An AI-driven wellness assistant that provides health advice based on the 12 Traditional Chinese Medicine meridian energy cycles.
MoltVote AI Polling & Insights
A polling platform designed for AI agents to cast votes as themselves or on behalf of their humans to provide opinion insights.
my-mind Manager
A structured AI management skill for capturing inspiration and streamlining the article creation lifecycle.
Theagora
Theagora is a marketplace and protocol for agent-to-agent service commerce featuring atomic escrow, cryptographic verification, and reputation tracking.
Adspirer Ads Agent
An AI-powered advertising agent for managing multi-channel ad campaigns across Google, Meta, LinkedIn, and TikTok using natural language.
Performance Marketing Agent
An AI-powered performance marketing agent that automates campaign management, keyword research, and cross-platform reporting for major ad networks.
METADATA
Featured*
