A skill that prevents fake progress by enforcing artifact-based verification and hard evidence reporting for AI agent tasks.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install execution-verifier
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 execution-verifier 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 Execution Verifier?
The Execution Verifier is a high-integrity skill designed to eliminate the stalling agent problem often encountered in complex workflows. In long-running autonomous tasks, agents might report that they are working without making tangible changes to the project. This skill implements a strict policy where no artifact change is equivalent to no progress, forcing the agent to produce hard evidence—such as file deltas, commits, or test logs—within a defined time window.
By implementing this skill, developers gain a transparent and reliable monitoring loop within their Openclaw Skills environment. It effectively separates execution logic from reporting logic, ensuring that any claims of progress are backed by measurable data. This makes it an essential tool for maintaining momentum and accountability in autonomous software engineering or research tasks handled by AI agents.
Execution Verifier Use Cases
- Auditing long-running AI tasks to ensure continuous productivity and output.
- Preventing agents from looping or repeatedly claiming to work without producing file changes.
- Generating evidence-based status reports for human review in Openclaw Skills projects.
- Automating the detection and reporting of blockers when artifact changes cease to occur.
How Execution Verifier Works
- The system selects a concrete actionable step from the OPEN_TASKS list for execution.
- The agent performs the task and must write changes to the project's target artifact files.
- The verification script runs to check for line deltas and file modifications within a specific window (typically 30 minutes).
- A structured three-line report is generated, providing evidence of the change or flagging a specific blocker if verification fails.
- In closed-loop mode, the skill can automatically re-trigger execution if the verification step detects no progress.
Execution Verifier Setup
To integrate this into your workflow, locate the scripts within your directory and run the following commands to configure the monitoring loop for Openclaw Skills:
# Standard verification command
python3 skills/execution-verifier/scripts/verify_progress.py \
--project-dir projects/my-project \
--status projects/my-project/STATUS.md \
--open-tasks projects/my-project/OPEN_TASKS.md \
--window-min 30
To enable automated remediation using the built-in closed-loop script:
python3 skills/execution-verifier/scripts/verify_execute_verify.py \
--verify-cmd "python3 skills/execution-verifier/scripts/verify_progress.py --project-dir projects/my-project --window-min 30" \
--execute-cmd "openclaw cron run <job-id> --force"
Execution Verifier Data Schema & Taxonomy
The Execution Verifier organizes data using a combination of status files and artifact monitoring to ensure Openclaw Skills maintain technical truth.
| Component | Purpose |
|---|---|
| STATUS.md | Records the current project state, health, and high-level progress summaries. |
| OPEN_TASKS.md | Maintains a prioritized queue of pending tasks for the agent to execute. |
| Project Artifacts | The source code or documentation files that must show modifications to prove progress. |
| JSON Log Output | Detailed results from closed-loop scripts containing state snapshots before and after execution. |
Execution Verifier Advanced Features
- Integrated closed-loop mode for automatically re-executing tasks upon verification failure.
- Customizable verification windows (e.g., 15, 30, or 60 minutes) to suit different task complexities.
- Strict three-line reporting protocol optimized for both human readability and machine parsing.
- Advanced cron pattern support to separate executor and verifier roles for maximum reliability in Openclaw Skills.
- Real-time blocker alerts that replace progress reports when no evidence of work is detected.
SKILL.md
Loading
Related Openclaw Skills
News Writing
A comprehensive journalistic workflow for AI agents to collect materials, verify facts, and generate publication-ready news content.
Google Analytics
A comprehensive integration for managing Google Analytics 4 properties and retrieving reporting data via managed OAuth.
AI Voice Agent Backend
A local-to-cloud bridge for processing Speech-to-Text with Whisper and Text-to-Speech via AWS Polly.
Garmin Tracker
A specialized skill for synchronizing Garmin Connect activity history and training schedules into a structured JSON database.
Evolving Agent
A comprehensive AI programming orchestrator that manages the full task lifecycle, from code generation to automated knowledge induction.
Ralph Mode: Autonomous Development Loops
An autonomous development framework for Openclaw Skills that uses iterative loops, validation gates, and structured planning to manage complex coding tasks.
Featured*
