A sophisticated learning and memory system that identifies failure patterns and proposes skill improvements via human-reviewed GitHub pull requests.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install memory-feedback
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 memory-feedback 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 Memory and Feedback?
The Memory and Feedback skill provides a robust learning infrastructure designed specifically for the Openclaw Skills ecosystem. It acts as a centralized intelligence layer that monitors the performance of all other active skills by logging every episode and failure. By maintaining a historical record of interactions, the system can move beyond reactive task execution into proactive optimization.
This skill implements a disciplined feedback loop where the agent identifies its own shortcomings based on real-world data. When a recurring pattern of failure or human correction reaches a specific confidence threshold, the skill generates a technical proposal to modify the underlying behavior logic. This ensures that Openclaw Skills evolve continuously, becoming more accurate and efficient the longer they are in operation.
Memory and Feedback Use Cases
- Automatically tracking agent failures and human corrections across a multi-skill environment.
- Identifying recurring intent misinterpretations in conversational AI interfaces to refine scope boundaries.
- Generating automated skill improvement proposals based on high-confidence performance data.
- Managing long-term agent memory storage and token budgets to prevent system bloat.
- Implementing a human-in-the-loop governance model for autonomous skill updates via GitHub.
How Memory and Feedback Works
- Every active skill in the environment uses internal hooks to log successful episodes and failures into a structured local directory.
- A nightly pattern scanning algorithm analyzes the last 30 days of logs to identify clusters of similar failures or corrections.
- The system calculates a confidence score for each pattern based on frequency, recency, and the presence of human feedback.
- For patterns exceeding the 0.85 confidence threshold, the skill generates a detailed proposal for a code or documentation change.
- The system creates a new branch and a GitHub Pull Request, allowing a human developer to review and merge the proposed improvement into the relevant Openclaw Skills definition.
Memory and Feedback Setup
To activate the learning loop for your Openclaw Skills, ensure the following environment variables and dependencies are configured:
# Define the central data directory
export OPENCLAW_DATA_DIR="/data"
# Ensure jq is installed for JSON processing
sudo apt install jq
# Optional: Set up GitHub integration for automated PRs
export GITHUB_TOKEN="ghp_your_personal_access_token"
# Install the GitHub CLI for seamless PR management
which gh || sudo apt install gh
Note: Core memory logging functions without a GitHub token, but the automated improvement workflow requires repository access.
Memory and Feedback Data Schema & Taxonomy
The skill organizes its learning data within a structured hierarchy to ensure efficient pattern detection and auditability:
| Data Type | Path | Description |
|---|---|---|
| Episodes | /data/memory/episodes/ |
JSON logs of successful actions, token usage, and session IDs. |
| Failures | /data/memory/failures/ |
Detailed records of what went wrong and expected outcomes. |
| Patterns | /data/memory/patterns/ |
Aggregated intelligence identifying recurring system behaviors. |
| Proposals | /data/memory/proposals/ |
Markdown-formatted suggestions for skill-file modifications. |
| Rate Limits | /data/memory/rate-limits/ |
Configuration and live counters for token and storage budgets. |
Memory and Feedback Advanced Features
- Automated GitHub PR generation including diffs, evidence summaries, and confidence scores.
- Sophisticated pattern detection algorithm with recency bias and weighted human-correction factors.
- Comprehensive rate-limiting and circuit breakers to protect token budgets and system storage.
- Nightly autonomous scanning that avoids consuming resources during peak business hours.
- Permanent rejection handling where the agent learns from closed PRs to never re-propose ignored changes.
SKILL.md
Loading
Related Openclaw Skills
Greek Individual Taxes
A comprehensive automation skill for processing Greek individual tax returns, E1 forms, and ENFIA property taxes.
Greek Financial Statements
A professional accounting skill for generating ΕΓΛΣ-compliant financial statements and VAT summaries with automated completeness validation.
Greek Document OCR
A specialized local OCR engine designed for high-precision recognition of Greek business documents, invoices, and handwritten materials.
Greek Compliance & AADE Integration
A specialized automation agent for managing Greek tax compliance, AADE TAXISnet submissions, and EFKA payroll processing.
OpenClaw Greek Accounting Meta-Skill
A comprehensive orchestration layer that routes business commands across 18 specialized skills to automate the full Greek accounting lifecycle.
System Health Check
A comprehensive diagnostic tool that validates file integrity, directory structures, and security permissions for your Openclaw Skills environment.
METADATA
Featured*
