A continuous learning system for AI agents that logs failures, corrections, and user preferences to MeiliSearch for rapid query recall and auto-distills them into human-readable memory files.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install learnings
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 learnings 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 Learnings Skill?
The Learnings Skill is a robust, self-improving memory architecture designed to elevate the capabilities of autonomous AI agents. By capturing developer feedback, system failures, and specific configuration patterns, this package ensures your AI system learns from its mistakes in real-time. Operating local search infrastructure powered by MeiliSearch, it serves as a central hub among Openclaw Skills to prevent repetitive errors, record development rules, and adapt dynamically to custom environment preferences without leaking sensitive data.
Through continuous log processing and secure credential handling, developers can maintain an organized knowledge base of lessons learned during agent-driven coding sessions. The skill intelligently distinguishes between operational failures, user preferences, and explicit guidelines, creating a tailored operating model for local development environments.
Learnings Skill Use Cases
- Preventing repetitive build failures by storing successful error-resolution patterns directly in a fast-recall local index.
- Automatically extracting developmental insights and architectural constraints from daily markdown files.
- Building a structured database of versioning rules, coding preferences, and security compliance guidelines.
- Restricting AI agents from repeating actions that trigger authentication, network, or permission errors.
How Learnings Skill Works
- Pre-Action Search: Before executing any command, the agent queries the MeiliSearch index using the current task context to discover relevant past learnings.
- Fix Application: If a matching error pattern or correction rule is returned, the agent applies the documented fix prior to taking action.
- Failure Capture: If an execution fails or a human correction is made, the details are immediately recorded with high priority.
- Automated Log Extraction: The agent periodically parses local daily notes to extract structured insights and saves them to the index.
- Human-Readable Synthesis: Raw, index-stored entries are distilled into a cleanly formatted markdown summary for developer visibility.
Learnings Skill Setup
Configure the required local environment parameters in ~/.openclaw/workspace/.env:
MEILI_HOST=http://127.0.0.1:7700
MEILI_KEY=your-master-key-here
Initialize the MeiliSearch index and perform everyday tasks using the core execution scripts:
# Log a new learning entry manually
bash scripts/log_learning.sh --category failure --what "deployment fails with auth token expiry" --fix "Refresh token before each deploy" --importance 0.9 --tags "deploy,auth,token"
# Query past memory logs prior to running an operation
bash scripts/search_learnings.sh "github push" 5
# Perform dry-run and apply extraction from recent workspace logs
bash scripts/auto_extract.sh --days 7 --apply
Learnings Skill Data Schema & Taxonomy
The system structures persistent metadata and active scripts across a designated workspace layout:
| File / Path | Purpose | Format / Schema |
|---|---|---|
LEARNINGS.md |
Auto-generated human-readable summary of lessons | Markdown document |
scripts/setup_index.sh |
Configures and initialises the local MeiliSearch index schema | Bash / Shell Script |
scripts/log_learning.sh |
Utility script to insert a new entry into the DB index | Bash / CLI Wrapper |
scripts/search_learnings.sh |
Queries local database for past context matches | Bash calling Curl |
scripts/auto_extract.sh |
Parses markdown logs and updates MeiliSearch entries | Shell + Python |
scripts/distill_learnings.sh |
Compiles raw database index down to LEARNINGS.md |
Shell + Python |
Data Classification Categories
failure: Identifies build, compilation, or execution errors.correction: Represents post-error resolutions or instructions.approach: Outlines strategic, optimal pathways for recurring operations.version_rule: Strict rules for tracking major, minor, and patch levels.preference: Developer habits, tool selections, and environmental options.recurring_failure: Repeated system flags requiring elevated attention.rule: Explicit security boundaries and rigid operational parameters.learning: Broader domain takeaways and code quality improvements.
Learnings Skill Advanced Features
- Privacy-First Data Filtering: Automatically strips out sensitive variables, credentials, bearer tokens, and private API keys before indexing.
- Safe Execution Defaults: Both
auto_extract.shanddistill_learnings.shdefault to run in dry-run safety mode, requiring the explicit--applyflag to commit updates. - Local-First Fast Recall Engine: Integrates high-performance local MeiliSearch instances rather than cloud-dependent vector DBs for immediate action context lookup.
- Secure Scratch File System: Generates encrypted/restricted temporary files using
mktempand sets strictchmod 600flags to prevent system leakage.
SKILL.md
Loading
Related Openclaw Skills
LYGO TruthLightEcho
A unique creative coding skill that transforms visual, mathematical, and conceptual Truth×Light metrics into rich, self-similar harmonic echo soundscapes.
Serpico: Whistleblowing, Integrity, and Institutional Reform Assistant
An AI-powered ethical guide and knowledge base designed to explore systemic corruption, institutional reform, and the psychological journey of whistleblowing based on Frank Serpico's real-life story.
The Outsiders: Capital Allocation & CEO Success Guide
An AI agent skill modeled on William Thorndike's book that empowers agents to deliver strategic business advice on capital allocation, decentralized operations, and rational value creation.
Agritech & Precision Farming
An AI-powered precision agriculture decision support skill that integrates satellite remote sensing (NDVI) and IoT sensor data to optimize crop growth, soil health, and pest management.
Contagious: Why Things Catch On Skill
An AI coding agent skill that implements Jonah Berger's STEPPS framework to engineer viral word-of-mouth marketing and social transmission into your content and products.
Refinement Loop (Claude Opus 4.8)
An iterative generation, evaluation, and revision workflow optimized for Claude Opus 4.8 to achieve high-quality outputs via structured self-critique and cost-efficient model routing.
Featured*
