An independent third-party audit system designed to verify task quality and prevent performance degradation in automated agent workflows.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install task-auditor
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 task-auditor 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 Task Auditor?
Task Auditor functions as a rigorous oversight layer for AI-driven automation within the Openclaw Skills ecosystem. By separating the executor (the agent doing the work) from the auditor (the agent verifying the work), it creates a checks-and-balances system that ensures every task meets predefined technical standards.
This skill is specifically designed to combat 'lazy' behavior in AI models by enforcing quantitative metrics such as log length, iteration depth, and content substance. It provides users with a reliable way to scale automation without sacrificing the quality of the final output.
Task Auditor Use Cases
- Validating the output quality of autonomous coding and research agents.
- Preventing AI agents from submitting low-effort or hallucinated results during long-running tasks.
- Generating professional audit trails for compliance and technical quality control.
- Automating the oversight role in multi-agent workflows to reduce the need for constant human monitoring.
How Task Auditor Works
- The executor agent (e.g., auto-iterator) submits its completed task, execution logs, and final reports.
- Task Auditor captures the submission and performs a multi-dimensional quantitative analysis.
- The skill calculates a score out of 100 based on log lines, iteration count, report length, and substantive content.
- A rating is assigned, ranging from Excellent (Pass) to Fail (Reject).
- Detailed audit reports and daily logs are generated, with low-quality scores automatically triggering alerts for the human supervisor.
Task Auditor Setup
To deploy the auditor within your environment, ensure it has read access to your task and memory directories.
# Enable the auditor within your agent workspace
openclaw add task-auditor
# Configure the scoring thresholds in your skill settings
# Default pass mark is set to 60/100
Task Auditor Data Schema & Taxonomy
The skill organizes its audit data into specific directories to maintain a clear trail of accountability:
| File Path | Description |
|---|---|
tasks/audits/*.audit.md |
Detailed reports for individual tasks, including scores and feedback. |
memory/alert-*.md |
High-priority quality alerts triggered by failed audits. |
memory/audit-YYYY-MM-DD.md |
Daily summary logs of all auditing activities performed. |
Task Auditor Advanced Features
- Multi-dimensional scoring system covering execution logs, iteration counts, and keyword analysis.
- Automated alert mechanism that flags low-quality tasks (score < 40) for immediate rejection.
- Anti-forgery protection using immutable timestamps and transparent log recording.
- Seamless integration with other Openclaw Skills to provide a unified quality assurance layer across diverse automation workflows.
SKILL.md
Loading
Related Openclaw Skills
Progress Reporter
An automated reporting system that generates detailed progress updates for AI agents every 10 minutes during task execution.
Rocket.Chat
A comprehensive integration for managing Rocket.Chat messaging, channel operations, and user administrative tasks through a RESTful API.
tg-miniapp
A comprehensive toolkit for building and debugging Telegram Mini Apps, resolving common safe area, layout, and API integration challenges.
Workspace Analyzer
A diagnostic utility that scans OpenClaw workspaces to identify maintenance needs, file bloat, and organizational issues for AI agents.
Identity Anchor
A cryptographic tool for AI agents to maintain and verify their unique identity across session resets and model switches.
1688 Distributor AI Automation
An automated AI-driven workflow for searching, selecting, and distributing products from 1688.com directly to your online shop.
Featured*
