A sophisticated multi-agent workflow that orchestrates specialized sub-agents to conduct deep, multi-perspective research and produce professional reports.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install researching-in-parallel
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 researching-in-parallel 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 Researching in Parallel?
Researching in Parallel is a high-performance skill designed for Openclaw Skills that automates deep-dive investigations. Instead of a single pass, it orchestrates a team of specialized sub-agents—covering breadth, critical analysis, and evidence gathering—to synthesize a comprehensive final report. This skill is ideal for moving beyond simple factual lookups into professional-grade briefing papers and literature reviews.
The system leverages a structured approach to ensure high-quality outputs with a unified bibliography, supporting iterative research where previous results can be fed back as inputs for future runs. By utilizing the Openclaw Skills framework, it manages token costs and elapsed time efficiently while ensuring every claim is backed by a verifiable evidence base.
Researching in Parallel Use Cases
- Conducting deep dives into complex technical, strategic, or academic topics.
- Generating balanced, multi-perspective overviews for executive briefings or orientation.
- Creating comprehensive research reports with structured, deduplicated bibliographies.
- Updating existing research reports with the latest developments or new analytical lenses.
- Synthesizing large volumes of web search results and uploaded documents into a single source of truth.
How Researching in Parallel Works
- Brief Clarification: The agent defines the research topic, angle, and required output depth (Brief, Standard, or Comprehensive) with the user.
- Source Assembly: Any starting-point files, URLs, or previous reports are organized into a reproducible research record.
- Prompt Configuration: Specialized task prompts are generated for three concurrent sub-agents: Breadth Sweep, Critical Lens, and Evidence Pass.
- Parallel Execution: The system spawns the sub-agents to perform parallel research, respecting concurrency limits and model assignments.
- Bibliography Compilation: All sources identified by sub-agents are consolidated, deduplicated, and formatted into a unified bibliography file.
- Final Synthesis: A dedicated synthesis or updater agent is spawned to weave the individual research threads into a cohesive final report.
- Delivery: The final report, bibliography, and source extracts are saved to the workspace for immediate use or further iteration.
Researching in Parallel Setup
To use this skill, ensure your environment supports the Openclaw Skills framework and has sessions_spawn enabled. This skill must be installed within the agent's workspace to allow sub-agents access to internal templates.
# Navigate to your skills directory
cd /workspace/skills/
# Clone or copy the researching-in-parallel folder
# Ensure skill-config.json is configured with your preferred LLMs
Prerequisites:
- OpenClaw default configuration (
maxSpawnDepth 1). - Web search and file read/write access.
- Browser and PDF tools are strongly recommended for full content retrieval.
Researching in Parallel Data Schema & Taxonomy
The skill organizes all research artifacts within a dedicated workspace directory to ensure a reproducible record of the Openclaw Skills run.
| File / Folder | Description |
|---|---|
sources-provided-*.md |
A markdown record of all initial input sources and URLs. |
bibliography-*.md |
A unified, deduplicated list of all sources including AI-generated summaries and BibTeX blocks. |
[TOPIC]-report.md |
The final synthesized research document. |
prompt-*.md |
The individual task prompts generated for each sub-agent. |
source-extracts/ |
(Optional) Cleaned text extracts of cited sources for verification. |
Researching in Parallel Advanced Features
- Iterative Research Loops: Use the outputs and bibliographies from one run as the provided sources for the next to achieve infinite depth.
- Role-Specific Model Assignment: Assign different LLMs to specific roles (e.g., a faster model for breadth, a more reasoning-heavy model for critique) via
skill-config.json. - Source Extraction: Automatically save cleaned extracts of every cited source to create a verifiable, offline evidence base.
- Unified Bibliographies: Automatically generates BibTeX blocks and identifies single-pass-only sources to highlight research gaps.
- Customizable Prompt Injection: Inject run-specific context into sub-agent prompts to prioritize specific angles or address identified gaps.
SKILL.md
Loading
Related Openclaw Skills
Bitrix24 Agent
A specialized AI skill for designing, implementing, and debugging secure Bitrix24 REST API integrations and automations.
Whoop Health Integration
A comprehensive tool for fetching and analyzing Whoop recovery, strain, sleep, and HRV data via the official Whoop API.
LinkedIn Automation Skill
A powerful CLI-driven skill for comprehensive LinkedIn automation, ranging from profile scraping to Sales Navigator outreach.
Planka CLI Skill
A powerful command-line interface to manage Planka Kanban boards and project workflows directly from your terminal.
Zerion API Web3 Analytics
An AI-powered connector for querying comprehensive blockchain wallet data, token prices, and DeFi analytics via the Zerion API.
Flash Redeem Knight
A browser automation skill that converts prepaid redemption links into ready-to-use pickup codes by automating store and option selection.
Featured*
