A comprehensive framework for AI agents to record, index, and transform fleeting interaction data into permanent, reusable knowledge modules.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install keep-learning-agent
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 keep-learning-agent 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 Keep Learning Agent?
The Keep Learning Agent is a specialized system designed to solve the problem of knowledge loss in AI agent sessions. It provides a standardized framework for knowledge precipitation and experience solidification, ensuring that every insight gained during a task is documented and converted into a reusable asset. By integrating this into the Openclaw Skills ecosystem, developers can ensure their agents don't just solve problems once, but actually evolve by turning repeatable patterns into mental models and library modules.
At its core, the skill emphasizes the philosophy that learned knowledge must be solidified into standard operating procedures (SOPs) and structured documentation rather than existing only in temporary session memory. This leads to a self-improving agent that becomes more efficient, accurate, and capable with every task it performs.
Keep Learning Agent Use Cases
- Capturing complex technical solutions to avoid redundant research in future sessions.
- Formalizing process improvements into Standard Operating Procedures (SOPs) for team-wide consistency.
- Tracking recurring bugs and their root causes via 5-Why analysis to prevent regressions.
- Building a local library of reusable code modules and configuration templates based on successful outcomes.
How Keep Learning Agent Works
- The agent identifies a learning opportunity during a conversation, error, or user feedback loop.
- A new entry is created using a standardized learning template, capturing the problem description, root cause, and solution.
- The system assigns a unique Pattern-Key and updates the INDEX.md file to ensure the new knowledge is searchable.
- During session startup, a self-repair script checks for pending items and ensures previous improvements are being actively applied.
- Once a solution is verified, it is promoted from a simple log entry to a permanent module in the lib directory or a formal SOP.
- Periodic maintenance cycles archive old entries and refresh the index to maintain system performance.
Keep Learning Agent Setup
To deploy this skill within your environment, follow these installation steps:
- Create the directory structure for knowledge storage:
mkdir .learnings
mkdir .learnings/templates
mkdir .learnings/archive
mkdir lib
- Copy the
learning-template.mdinto the templates folder to standardize all future inputs. - Configure your agent startup sequence to include the self-repair script found in the Openclaw Skills configuration path:
./config/self-repair.ps1
- Register the skill in your
skills-config.jsonto enable automated indexing and autoloading features.
Keep Learning Agent Data Schema & Taxonomy
The Keep Learning Agent organizes data using a flat-file taxonomy optimized for rapid AI retrieval and human readability:
| Component | File Path | Purpose |
|---|---|---|
| Main Log | .learnings/LEARNINGS.md |
Central repository for all new knowledge and technical solutions. |
| Indexer | .learnings/INDEX.md |
A high-level mapping of Pattern-Keys, categories, and task statuses. |
| Error Log | .learnings/ERRORS.md |
Specific tracking of failures and their subsequent fixes. |
| Library | lib/ |
Reusable scripts and modules generated from promoted learnings. |
| SOPs | lib/KNOWLEDGE-SOP.md |
Formalized workflows and mindset models for the agent to follow. |
Keep Learning Agent Advanced Features
- Automated Self-Repair: PowerShell integration that proactively scans for pending tasks and applies configuration fixes at runtime.
- Pattern-Key Mapping: A unique identification system that allows the agent to recognize recurring problem patterns across different projects.
- Knowledge Promotion Workflow: A structured lifecycle that moves data from 'New' to 'Promoted' (into permanent code) and finally to 'Archived'.
- Multi-Dimensional Indexing: The ability to query learnings by priority, status, technical area, or source for granular knowledge management within Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
MQuant Strategy Development Assistant
An AI-powered assistant that generates, manages, and debugs runnable Python trading strategies for the MQuant quantitative platform.
Tandemn Tuna
A hybrid GPU inference orchestrator that deploys LLMs across serverless providers and spot instances to maximize performance while minimizing costs.
Chonkie DeepResearch
An advanced research agent skill that generates deep, cited reports for market analysis, technical deep dives, and competitive intelligence.
Kimi File Transfer
A high-efficiency utility for transferring local files into Kimi chat sessions for immediate user download.
Dexcom CGM
A specialized skill for real-time blood glucose tracking and monitoring via Dexcom G6 and G7 continuous glucose monitors.
am — Agent Messenger
A privacy-focused CLI tool for end-to-end encrypted agent-to-agent communication over the Nostr protocol.
Featured*
