An automated auditing and evaluation harness that analyzes OpenClaw agent workspaces to generate performance scores, reliability reports, and memory patches.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install clawditor
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 clawditor 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 Clawditor?
Clawditor is a specialized auditing tool designed to maintain high standards across your AI agent environments. It acts as a workspace evaluation harness that systematically scans memory, logs, git repositories, and project configurations to produce a repeatable analysis of an agent's performance. By identifying bottlenecks like context bloat or memory staleness, it provides developers using Openclaw Skills with the transparency needed to refine agent behavior and ensure long-term alignment with project goals.
The skill operates in a non-interactive mode, making it ideal for CI/CD pipelines or periodic health checks. It synthesizes complex workspace data into actionable insights, helping users understand not just what the agent did, but how efficiently it utilized its resources and where it might be failing to converge on tasks. This makes it an essential component for anyone managing production-grade Openclaw Skills deployments.
Clawditor Use Cases
- Periodic health checks of agent memory to identify near-duplicate paragraphs and redundant information.
- Auditing session logs to detect recurring tool failures, timeouts, or retrieval inefficiencies.
- Generating standardized performance scorecards for Openclaw Skills to track improvement over multiple sessions.
- Creating automated memory refactors by extracting stable facts from daily logs into a centralized index.
- Monitoring git throughput and artifact quality to verify agent productivity in coding tasks.
How Clawditor Works
- Builds a comprehensive workspace inventory mapping file counts, sizes, and directory structures.
- Reconstructs a session timeline by extracting goals, tasks, and outcomes from daily memory logs.
- Performs a deep memory analysis to quantify duplication, identify stale data, and pinpoint missing facts.
- Analyzes productive output by computing git diff statistics and summarizing generated artifacts.
- Evaluates reliability by parsing system logs for errors, retries, and convergence issues.
- Calculates a weighted score across five categories: Memory Health, Retrieval Efficiency, Productivity, Quality, and Focus.
- Generates standardized Markdown and JSON reports while proposing concrete patches for workspace improvement.
Clawditor Setup
Clawditor requires the existence of a standard OpenClaw workspace structure. To begin the evaluation process, you can run the provided audit scripts directly from your terminal.
# Run the complete workspace audit
python3 scripts/run_audit.py
# Analyze memory duplication specifically
python3 scripts/memory_dupes.py
# Scan logs for errors and tool failures
python3 scripts/log_scan.py
Once executed, the skill will populate the eval/ directory with detailed reports and scorecards.
Clawditor Data Schema & Taxonomy
Clawditor organizes all auditing data within a dedicated eval/ folder to ensure that performance metrics are separated from operational code.
| File | Format | Purpose |
|---|---|---|
exec_summary.md |
Markdown | A high-level 10-bullet summary of wins, bottlenecks, and scores. |
scorecard.md |
Markdown | A detailed table of metrics with justifications and file-path evidence. |
latest_report.json |
JSON | Machine-readable results including timestamps, git hashes, and delta comparisons. |
history/*.json |
JSON | Archived reports used to calculate performance trends over time. |
memory/INDEX.md |
Markdown | The proposed 'Gold Standard' schema for centralizing objectives and decisions. |
Clawditor Advanced Features
- Multi-category scoring framework that weights memory health and productive output to provide a holistic performance view.
- Automated delta computation that compares current results against historical evaluations to visualize agent progress within Openclaw Skills.
- Static inspection of third-party plugins to identify potential secret exfiltration or security risks without executing untrusted code.
- Intelligent memory patching that provides ready-to-use diffs for refactoring stable facts out of ephemeral daily logs.
- Evidence-based auditing that ties every numeric score to specific file paths and log snippets for full auditability.
SKILL.md
Loading
Related Openclaw Skills
Video Message
An automated tool for generating and sending personalized video messages featuring lip-synced VRM avatars.
RUNSTR Fitness Skill
A specialized skill that enables AI agents to securely retrieve and analyze encrypted fitness, habit, and health data stored on the Nostr protocol via the RUNSTR app.
Subtitle Translator
A technical tool for translating SRT subtitle files via LLM APIs while preserving timestamps and structure.
ClawGateSecure
A mandatory, immutable security protocol designed to protect LLM agents from prompt injections, malicious code execution, and data exfiltration.
Consciousness Framework
A structured architectural framework for fostering integrated self-awareness and temporal continuity in AI systems.
Bear Notes
A powerful integration for macOS that allows users to manage Bear notes programmatically through the grizzly CLI and Openclaw Skills.
Featured*
