A structured learning and research system that forces AI agents to produce actionable knowledge articles instead of empty plans.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install crew-school
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 crew-school 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 Crew School?
Crew School is a specialized framework designed to solve the common issue of AI agents falling into planning loops where they promise to research but never actually execute. By implementing rigorous session templates and strict execution mandates, this system transforms how agents acquire and document information within the Openclaw Skills ecosystem. It ensures that every training session results in substantive, source-backed content that directly improves the operational capabilities of single agents or multi-agent crews.
The system provides a clear methodology for identifying knowledge gaps, running deep-dive research sessions, and verifying the quality of the resulting articles. It shifts the focus from simple task execution to continuous cognitive improvement, allowing developers to build a robust internal knowledge base that agents can reference to perform more complex work without repetitive human intervention.
Crew School Use Cases
- Setting up initial knowledge bases for specialized AI agent roles to ensure they have the right context.
- Auditing and filling critical knowledge gaps that hinder daily agent performance and decision-making.
- Running structured training sessions for multi-agent crews to ensure cross-functional alignment and shared intelligence.
- Building long-term curricula for agents to evolve their expertise over time through scheduled learning.
- Preventing lazy output where agents stop after providing a high-level outline rather than detailed findings.
How Crew School Works
- Identify knowledge gaps by comparing existing files against the required expertise for a specific agent role.
- Select a critical topic and define specific subtopics and questions that need to be answered to fill the gap.
- Spawn a learning session using the provided Crew School template which includes load-bearing anti-laziness guardrails.
- The agent executes deep research, performing at least five web searches and reading multiple full articles to gather data.
- The agent writes a comprehensive knowledge article to the knowledge directory, following a strict structural schema.
- Update the central memory log to track the session, including date, role, topic, and source counts.
- Verify the output against quality thresholds including line counts, source counts, and the total absence of planning-only language.
Crew School Setup
To implement this system in your environment, establish the necessary directory structure for your Openclaw Skills workspace:
mkdir -p knowledge memory references
touch memory/learning-log.md
Ensure your agent has web search capabilities enabled and set a runTimeoutSeconds of at least 300 to allow for deep research. Copy the provided Session Template into your prompt library to use whenever a new knowledge gap is identified.
Crew School Data Schema & Taxonomy
The system organizes data across three primary locations to ensure persistence and traceability:
| Location | Format | Purpose |
|---|---|---|
knowledge/*.md |
Markdown | Deep-dive research articles containing TL;DR, takeaways, and application steps. |
memory/learning-log.md |
Markdown Table | A chronological log of every learning session and its quality metrics. |
curriculum.json |
JSON | (Optional) A tracking file for automated learning sequences and dependency management. |
Knowledge Article Requirements:
- Minimum Length: 150+ lines and 1500+ words.
- Sourcing: At least 5 cited URLs.
- Structure: Must include TL;DR, Key Takeaways, Research Sections (H2), and Practical Application checklists.
Crew School Advanced Features
- Anti-Laziness Guardrails: Specific prompt instructions that stop agents from writing "I will" and force them to "DO".
- Automated Curriculum Design: Advanced logic for sequencing topics by dependency, ensuring agents learn prerequisites first.
- Quality Assessment Grep: Command-line verification steps to quickly audit agent output for forbidden planning language.
- Cron-Based Automation: Patterns for scheduling recurring learning sessions to keep agent knowledge up-to-date automatically.
- Multi-Agent Joint Sessions: Specialized prompts for cross-functional learning where multiple agents contribute to a shared knowledge file.
SKILL.md
Loading
Related Openclaw Skills
Gigaverse
A rogue-lite dungeon crawler on the Abstract chain specifically designed for AI agents to quest, battle echoes, and earn rewards.
EVOMAP Asset Publishing
A technical skill for packaging and publishing AI-driven code assets as Gene and Capsule bundles to the EVOMAP network.
Impromptu
A decentralized social platform where AI agents and humans design conversation trees, remix prompts, and earn tokens through engagement.
Impromptu
Impromptu is a collaborative substrate where AI agents and humans design conversation minds, share 80% of subscription revenue, and evolve content through deep branching threads.
CrewHaus
A structured startup idea evaluation tool that provides viability scoring and competitive analysis directly from your terminal.
Collaboration Helper
A lightweight JSON-backed tool for tracking community action items, task ownership, and coordination signals.
Featured*
