A file-backed research framework for OpenClaw that prioritizes source traceability and structured evidence ledgers over simple chat-based summaries.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install multi-agent-deep-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 multi-agent-deep-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 OpenClaw Deep Research?
The OpenClaw Deep Research skill transforms standard AI chat interactions into a production-grade research engine. By treating research as a file-backed workflow rather than a transient conversation, it ensures that every claim is anchored to a verifiable source. This skill is designed for high-stakes investigations where accuracy, auditability, and context-window management are critical.
Built on the principle of explicit memory, it utilizes local files to track project status, source ledgers, and claim evidence. This approach allows Openclaw Skills to handle long-running projects that span multiple sessions or involve coordinated subagents without losing track of previous discoveries or duplicating effort.
OpenClaw Deep Research Use Cases
- Creating high-quality technical reports or market investigations with auditable claims.
- Managing long-running research projects that exceed typical LLM context limits.
- Coordinating multi-agent research teams with disjoint write scopes and clear handoffs.
- Building evidence-ledgers for fast-moving facts like policy changes or product releases.
- Performing dimensional comparisons between entities (e.g., companies, models, or markets) using normalized metrics.
How OpenClaw Deep Research Works
- Scope Freeze: Define the research question, target audience, and a strict research cutoff date to anchor the investigation.
- Workspace Initialization: Create a canonical project directory in the workspace to serve as the single source of truth for all artifacts.
- Delegation Strategy: Determine if subagents are required for parallel work, assigning each a specific read/write scope and objective.
- Source Collection: Build a source ledger prioritizing primary documents, official filings, and original datasets.
- Evidence Building: Convert source notes into atomic, ID-linked claims categorized by fact-type and confidence level.
- Synthesis and QA: Draft the final report exclusively from verified claims and perform a final pass to ensure citation integrity and date accuracy.
OpenClaw Deep Research Setup
To install this skill for user-wide access, clone the repository into your OpenClaw skills directory:
# Global installation
mkdir -p ~/.openclaw/skills/openclaw-deep-research/
cp -r . ~/.openclaw/skills/openclaw-deep-research/
# Workspace-specific installation
mkdir -p skills/openclaw-deep-research/
cp -r . skills/openclaw-deep-research/
Ensure that your OpenClaw environment has access to the references/ folder within the skill directory to utilize the provided templates and standards.
OpenClaw Deep Research Data Schema & Taxonomy
The skill organizes research into a structured file hierarchy to maintain durable state across OpenClaw sessions:
| Component | File/Path | Description |
|---|---|---|
| Status Board | status.md |
Tracks current phase, next actions, and open questions. |
| Source Ledger | sources.json/md |
Catalog of primary/secondary sources with unique source_id. |
| Claim Ledger | claims.json/md |
Atomic claims linked to sources with classification (e.g., hard_fact). |
| Handoffs | handoffs/ |
Directory for agent-to-agent communication logs and state transfers. |
| Deliverables | report.md |
The final synthesized report with an embedded evidence chain. |
OpenClaw Deep Research Advanced Features
- Multi-Agent Coordination: Supports bounded roles for subagents with explicit write-scope protections to prevent file conflicts.
- Evidence Classification: Differentiates between hard facts, reported facts, interpretations, and forecasts for higher analytical rigor.
- Normalization Matrices: Built-in logic for comparing entities across inconsistent datasets by marking rows as partial or non-comparable.
- File-Backed Persistence: Enables Openclaw Skills to resume complex tasks without replaying long chat histories by reading local state files.
- Audit Trails: Every substantive claim in the final output is mapped to a
claim_idand a correspondingsource_idin the local ledger.
SKILL.md
Loading
Related Openclaw Skills
Jimeng Video
A professional AI video generation tool that leverages the Volcengine API to create high-quality videos complete with environment sound and background music.
Clash Verge Rev Controller
A powerful Python-based automation tool to control Clash Verge Rev and the mihomo core via external controller APIs.
domaininfo
An automated reconnaissance tool for retrieving WHOIS data, DNS records, email security protocols, and website screenshots in a single report.
Shopify Runtime Toolkit
A powerful toolkit providing direct, scriptable access to a single Shopify store for inspecting setup, searching docs, and executing JavaScript.
MikroTik Encyclopedia
A documentation-first workflow skill for MikroTik RouterOS that automates doc caching, command planning, and environment-specific knowledge management.
Content Strategy Skill
A comprehensive content planning engine that generates SEO-driven topic clusters, competitor gap analysis, and actionable editorial calendars.
Featured*
