1. Prerequisites
- The full OpenClaw main project has been deployed and all services are running normally.
- Git command-line tool installed.
- GitHub App created with read/write, Issue, and PR permissions granted to target repositories.
- Valid LLM API Key (OpenAI / Qwen / Claude, etc.).
- Server permissions to edit files and restart services.
2. Pull Latest Skill Source Code
Navigate to the root directory of your OpenClaw project and pull the latest remote code to synchronize the ClawSweeper skill files.
bash
cd /your-openclaw-root-folder
git pull origin main
Verify the skill directory exists:
.agents/skills/clawsweeper3. Generate and Modify Skill Configuration File
Step 1: Enter the skill folder
bash
cd .agents/skills/clawsweeper
Step 2: Copy the template config to create an active config file
bash
cp config.example.yaml config.yaml
Step 3: Edit config.yaml and fill in core parameters
yaml
github_app_id: "Your GitHub App ID"
github_private_key_path: "Full file path to your App private key"
target_repo: "repository-owner/repository-name"
scan_interval: 10 # Scheduled scan interval in minutes
protect_labels: ["bug-critical", "hold"] # Issues with these labels will not be auto-closed
auto_close: true # Enable automatic issue archiving
llm_model: "gpt-4o" # Name of your selected LLM model
Step 4: Configure global LLM API credentials
Edit the root
.env file in the project root and add the LLM endpoint URL and API key:env
LLM_API_KEY=your-llm-api-key
LLM_BASE_URL=llm-api-base-endpoint-address
4. Enable the Skill Toggle
Open the skill registration config inside the clawsweeper folder and set the enable flag to
true:yaml
enable: true
Save the file.
5. Restart OpenClaw to Load the New Skill
Option A: Docker Compose Deployment
bash
# Stop running containers
docker compose down
# Restart services in detached background mode
docker compose up -d
Option B: Native Binary Deployment
bash
# Terminate existing agent processes
pkill -f openclaw
# Launch agent service in background
nohup ./openclaw agent run &
6. Verify Loading Status & Troubleshoot Logs
Stream real-time logs to confirm ClawSweeper loads without errors:
bash
# For Docker Compose environments
docker compose logs -f openclaw | grep ClawSweeper
# For native binary environments
tail -f agent.log | grep ClawSweeper
The message
ClawSweeper skill loaded successfully confirms successful initialization.7. Functional Testing
Navigate to the comment section of any Issue in your target GitHub repository and send the following test commands:
- Check skill running status
plaintext
/clawsweeper status
- Manually trigger a full repository Issue scan
plaintext
/clawsweeper scan
- View current auto-close rule configuration
plaintext
/clawsweeper rule
8. Common Troubleshooting Solutions
- GitHub authorization failure
Check private key file path, GitHub App permission scopes, and repository access authorization.
- LLM API calling errors
Double-check the API key, base endpoint URL, and model name set in the
.envfile. - Skill not taking effect
Confirm
enable: trueis set, restart the service, and verify all skill folders and files are intact. - Issues not auto-closing
Check for conflicting labels under
protect_labelsand confirm theauto_closetoggle is enabled.
