A persistent memory system that logs agent corrections and automatically injects them into future sessions to prevent recurring errors.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install correction-memory
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 correction-memory 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 Correction Memory?
Correction Memory is a specialized utility for Openclaw Skills designed to solve the persistent issue of agents repeating the same mistakes across different sessions. Typically, when an agent's output is corrected, that knowledge is lost once the session ends. This skill bridges that gap by providing a dedicated logging and injection mechanism that captures feedback and applies it to every future spawn of the same agent type.
By integrating the correction-tracker library into your workflow, you transform ephemeral corrections into a permanent knowledge base. Whether you are enforcing specific coding standards or stylistic preferences, this skill ensures your agents evolve and improve over time rather than resetting to a baseline state with every new task.
Correction Memory Use Cases
- Eliminating repetitive formatting or syntax errors in generated code.
- Enforcing project-specific architectural patterns that agents frequently ignore.
- Training creative writing agents to maintain consistent style and tone across multiple chapters.
- Reducing manual prompting overhead by automating the delivery of previous feedback.
How Correction Memory Works
- The system captures corrections via programmatic calls or natural language instructions provided to the main agent.
- Corrections are logged into structured .jsonl files categorized by agent type within the memory directory.
- When a subagent is spawned, the agent-context-loader identifies the agent type (e.g., CoderAgent) from the task description.
- The system queries the correction log for entries associated with that agent type from the last 30 days.
- A correction preamble is generated and prepended to the agent's prompt, ensuring past lessons are top-of-mind for the current task.
Correction Memory Setup
To install this capability for your Openclaw Skills setup, follow these steps:
Step 1: Install the Tracker
Copy the tracker template to your local workspace:
cp references/correction-tracker-template.js $OPENCLAW_WORKSPACE/lib/correction-tracker.js
Verify the installation:
node $OPENCLAW_WORKSPACE/lib/correction-tracker.js
Step 2: Integration
If you are already using the intent-engineering skill, the injection hook into agent-context-loader.js is automatic. If not, you must manually import buildCorrectionPreamble from the tracker library into your custom spawn logic.
Correction Memory Data Schema & Taxonomy
The skill maintains a flat-file database using JSONL for high performance and transparency. Data is stored in $OPENCLAW_WORKSPACE/memory/corrections/[AgentType].jsonl.
| Field | Description |
|---|---|
agent_type |
The category of agent receiving the correction (e.g., CoderAgent, AuthorAgent) |
issue |
A description of the incorrect behavior observed |
correction |
The specific rule or instruction to follow in the future |
timestamp |
Used to manage the 30-day auto-expiration of stale rules |
metadata |
Optional context such as session ID or communication channel |
Correction Memory Advanced Features
- Auto-expiration logic that removes corrections older than 30 days to prevent prompt bloat.
- Natural language processing that allows you to log corrections by simply telling the main agent what went wrong.
- Keyword-based agent type detection that automatically routes corrections to the correct library.
- Seamless interoperability with other Openclaw Skills to create a cohesive autonomous agent environment.
SKILL.md
Loading
Related Openclaw Skills
Agent Stability Framework (ASF)
A comprehensive three-layer defense system to prevent AI personality drift, technical faults, and soul misalignment in autonomous agents.
Video Watcher
A specialized tool for programmatically fetching and reading transcripts from YouTube and Bilibili videos to enable AI-powered analysis.
Joan Workflow
A workspace-based knowledge and task management system designed specifically for AI-assisted development workflows.
Corpus
A powerful integration for AI agents to interact with the Corpus AI library for content retrieval, web bookmarking, and reminder management.
Intent Engineering
A machine-readable intent layer that ensures subagents optimize for your actual priorities rather than just completion metrics.
Cost Governor
A pre-flight cost estimation and approval gate system designed to prevent API overspend during multi-agent deployments.
METADATA
Featured*
