TL;DR
OpenClaw uses Tools and Skills to take action. Tools are low-level capabilities such as reading files, running commands, and searching the web. Skills are pre-built workflows that combine multiple tools for tasks such as code review and meeting summaries. This guide covers the 25 default tools, 53+ community skills, tool and skill management, and custom skill creation.
Tools vs. Skills
OpenClaw combines two systems:
- Tools define what OpenClaw can do.
- Skills define how OpenClaw uses those capabilities in a workflow.
A tool might read a file or execute a command. A skill combines several tools to complete a larger task, such as reviewing a pull request.
Example: Code Review
With tools only, OpenClaw may be able to read files but not know how to perform a complete review.
A code_review skill can orchestrate the process:
- Read changed files with
read. - Search for related code with
search. - Check documentation or similar issues with
web_search. - Post review comments with
github.
Skills provide the workflow; tools provide the individual capabilities.
The 25 Default Tools
Tools are built into OpenClaw. Enable only the tools you trust and need.
File System Tools
1. read
Reads file contents.
Use cases:
- View source code.
- Inspect configuration files.
- Analyze logs.
Security: Read-only; it cannot modify files.
Example:
You: Read package.json
OpenClaw: [Shows file contents]
2. write
Creates or modifies files.
Use cases:
- Generate code.
- Update configuration.
- Create documents.
Security: Can overwrite files, so use it carefully.
Example:
You: Create test.txt with "Hello World"
OpenClaw: Done. Created test.txt
3. list
Lists files and directories.
Use cases:
- Explore a project structure.
- Find files.
- Inspect source directories.
Security: Read-only.
Example:
You: List files in src/
OpenClaw:
- api/
- components/
- utils/
- index.js
4. search
Searches file contents, similar to grep.
Use cases:
- Find code patterns.
- Locate functions.
- Search for TODOs or configuration values.
Security: Read-only.
Example:
You: Search for "TODO" in all JavaScript files
OpenClaw:
- src/api/users.js:45
- src/utils/helpers.js:23
...
Execution Tools
5. exec
Runs shell commands.
Use cases:
- Run tests.
- Build projects.
- Deploy code.
Security: High risk. It can execute arbitrary commands, including destructive commands such as rm -rf.
Example:
You: Run npm test
OpenClaw: [Executes tests and shows results]
Enable exec only when you fully trust the environment and OpenClaw's permissions.
6. python
Executes Python code.
Use cases:
- Data analysis.
- Scripting.
- Automation.
Security: Can run arbitrary Python code.
Example:
You: Calculate the average of [10, 20, 30, 40]
OpenClaw: Average: 25
7. node
Executes Node.js code.
Use cases:
- JavaScript scripting.
- API calls.
- Node-based automation.
Security: Can run arbitrary JavaScript.
Example:
You: Fetch data from https://api.example.com/users
OpenClaw: Retrieved 50 users
Web Tools
8. web_search
Searches the internet, including Google and Bing.
Use cases:
- Research.
- Fact-checking.
- Finding documentation.
Security: Low risk; read-only.
Example:
You: Search for OpenClaw documentation
OpenClaw:
1. docs.openclaw.ai - Official docs
2. github.com/openclaw - Repository
...
9. web_fetch
Downloads web page content.
Use cases:
- Read articles.
- Scrape data.
- Monitor websites.
Security: Low risk; read-only.
Example:
You: Fetch https://example.com
OpenClaw: [Downloads and shows page content]
10. web_screenshot
Captures screenshots of web pages.
Use cases:
- Visual testing.
- Website monitoring.
- Documentation.
Security: Low risk.
Example:
You: Screenshot https://myapp.com
OpenClaw: Saved screenshot to screenshots/myapp.png
Memory Tools
11. memory
Stores and retrieves information across sessions.
Use cases:
- Remember preferences.
- Save useful context.
- Reuse project decisions.
Security: Low risk; uses local storage only.
Example:
You: Remember that I prefer TypeScript over JavaScript
OpenClaw: Saved to memory.
Later:
You: Generate a new API endpoint
OpenClaw: Creating a TypeScript endpoint...
12. context
Manages conversation context.
Use cases:
- Track multi-step tasks.
- Maintain state.
- Keep related requests connected.
Security: Low risk.
Example:
You: I'm working on the user authentication feature
OpenClaw: Noted. I'll keep this context for our conversation.
You: Add password validation
OpenClaw: Adding validation to the auth feature...
Scheduling Tools
13. schedule
Creates recurring tasks, similar to cron jobs.
Use cases:
- Daily reports.
- Monitoring.
- Reminders.
Security: Medium risk because tasks run automatically.
Example:
You: Check Hacker News every morning at 8 AM and send me the top 3 posts
OpenClaw: Scheduled. I'll run this daily at 8 AM.
14. heartbeat
Runs background tasks at intervals.
Use cases:
- Monitoring.
- Polling.
- Periodic checks.
Security: Medium risk.
Example:
You: Monitor https://api.myapp.com every 5 minutes
OpenClaw: Heartbeat started. Checking every 5 minutes.
Communication Tools
15. email
Sends emails.
Use cases:
- Notifications.
- Reports.
- Alerts.
Security: Medium risk; it can send email on your behalf.
Example:
You: Email the team a summary of today's deployments
OpenClaw: Sent email to team@company.com
16. slack
Posts messages to Slack channels.
Use cases:
- Team notifications.
- Alerts.
- Deployment updates.
Security: Medium risk.
Example:
You: Post to #engineering: "Deployment complete"
OpenClaw: Posted to #engineering
17. discord
Sends Discord messages.
Use cases:
- Community updates.
- Bot interactions.
- Operational notifications.
Security: Medium risk.
Example:
You: Send a message to #general: "Server maintenance in 10 minutes"
OpenClaw: Message sent to #general
Integration Tools
18. github
Interacts with GitHub repositories, pull requests, and issues.
Use cases:
- Code reviews.
- Issue management.
- Deployment workflows.
Security: Medium risk; it can modify repositories.
Example:
You: Create an issue: "Fix login bug on mobile"
OpenClaw: Created issue #456 in myapp repo
19. jira
Manages Jira tasks.
Use cases:
- Project management.
- Task tracking.
- Finding assigned work.
Security: Medium risk.
Example:
You: What tasks are assigned to me?
OpenClaw:
- PROJ-123: Fix authentication
- PROJ-124: Update docs
- PROJ-125: Code review
20. database
Queries SQL and MongoDB databases.
Use cases:
- Data analysis.
- Reporting.
- Debugging.
Security: High risk because it can modify data.
Example:
You: How many users signed up today?
OpenClaw: 47 new users today
Utility Tools
21. calculator
Performs calculations, conversions, and estimates.
Security: No risk.
Example:
You: Calculate 15% of $1,250
OpenClaw: $187.50
22. date_time
Handles dates, times, time zones, and time calculations.
Security: No risk.
Example:
You: What time is it in Tokyo?
OpenClaw: 3:45 AM JST (March 6, 2026)
23. image_gen
Generates images using services such as DALL-E or Stable Diffusion.
Use cases:
- Thumbnails.
- Mockups.
- Illustrations.
Security: Low risk, although image generation may use API credits.
Example:
You: Generate an image of a robot using a computer
OpenClaw: Saved to images/robot.png
24. pdf
Creates and reads PDF files.
Use cases:
- Reports.
- Documentation.
- Invoices.
Security: Low risk.
Example:
You: Convert report.md to PDF
OpenClaw: Created report.pdf
25. zip
Compresses and extracts files.
Use cases:
- Backups.
- File transfers.
- Archives.
Security: Low risk.
Example:
You: Zip the src/ directory
OpenClaw: Created src.zip (2.3 MB)
Managing Tools
View Enabled Tools
Ask OpenClaw which tools are currently enabled:
You: What tools do I have enabled?
OpenClaw:
Enabled:
- read, write, list, search
- web_search, web_fetch
- memory, context
Disabled:
- exec
- database
Enable a Tool
Enable tools only when a workflow requires them:
You: Enable the exec tool
OpenClaw: Warning: exec allows running shell commands. This is powerful but risky. Are you sure?
You: Yes
OpenClaw: Enabled exec tool
Disable a Tool
Remove access when a tool is no longer needed:
You: Disable the database tool
OpenClaw: Disabled database tool. I can no longer query databases.
Tool Security Checklist
Start with the smallest useful set of tools.
High-risk tools:
-
exec: Can run arbitrary commands. -
database: Can modify data. -
email: Can send messages on your behalf.
Enable these only when you trust OpenClaw and have reviewed the workflow.
Medium-risk tools:
-
write: Can overwrite files. -
github: Can modify repositories. -
slack: Can post messages.
Enable these for automation, but monitor their usage.
Lower-risk tools:
-
read,list,search: Read-only file operations. -
web_search,web_fetch: Read-only web operations. -
memory: Local storage.
These are generally safer defaults.
The 53+ Community Skills
Skills are pre-built workflows for common tasks. The community has created more than 53 skills.
Development Skills
code_review
Reviews code for quality, bugs, and style.
Tools used: read, search, web_search, github
You: Review PR #42
OpenClaw: [Analyzes code, checks for issues, and posts comments]
debug_assistant
Helps investigate and debug errors.
Tools used: read, search, exec, web_search
You: Debug this error: TypeError: Cannot read property 'name' of undefined
OpenClaw: [Analyzes the code and suggests fixes]
test_generator
Generates unit tests.
Tools used: read, write, exec
You: Generate tests for src/api/users.js
OpenClaw: [Creates a test file with comprehensive tests]
refactor_assistant
Suggests and applies code improvements.
Tools used: read, write, search
You: Refactor src/utils/helpers.js
OpenClaw: [Suggests improvements and applies changes]
Productivity Skills
meeting_summary
Summarizes meetings and extracts action items.
Tools used: memory, context
You: [Forward a meeting transcript]
OpenClaw: [Generates a structured summary with action items]
task_manager
Manages tasks across tools such as Jira, Asana, and Trello.
Tools used: jira, memory, schedule
You: What's on my plate this week?
OpenClaw: [Lists tasks from connected tools]
email_assistant
Drafts and sends emails.
Tools used: email, memory
You: Draft an email to the team about tomorrow's deployment
OpenClaw: [Generates a professional email]
Content Skills
blog_writer
Writes blog posts.
Tools used: web_search, write, memory
You: Write a blog post about OpenClaw use cases
OpenClaw: [Researches the topic and writes the post]
social_media_manager
Creates social media content.
Tools used: web_search, memory, slack
You: Create a Twitter thread about our new feature
OpenClaw: [Generates the thread]
video_script_writer
Writes video scripts.
Tools used: web_search, write, memory
You: Write a 10-minute YouTube script about OpenClaw setup
OpenClaw: [Generates a structured script]
Monitoring Skills
api_monitor
Monitors API health.
Tools used: web_fetch, heartbeat, slack
You: Monitor https://api.myapp.com
OpenClaw: [Checks the API every 5 minutes and alerts on issues]
error_tracker
Tracks and analyzes errors.
Tools used: read, search, memory, slack
You: Monitor error logs and alert me on new errors
OpenClaw: [Watches logs and sends alerts]
deployment_watcher
Monitors deployments.
Tools used: github, exec, slack
You: Watch for new deployments and run smoke tests
OpenClaw: [Monitors, tests, and reports]
Installing Skills
Browse Available Skills
List available skills before installing anything:
You: List available skills
OpenClaw:
Development:
- code_review
- debug_assistant
- test_generator
- refactor_assistant
Productivity:
- meeting_summary
- task_manager
- email_assistant
Content:
- blog_writer
- social_media_manager
- video_script_writer
Monitoring:
- api_monitor
- error_tracker
- deployment_watcher
Install a Skill
Install a skill by name:
You: Install skill code_review
OpenClaw:
Installing code_review skill...
Required tools:
- read: enabled
- search: enabled
- web_search: enabled
- github: not enabled
Enable github tool? (y/n)
You: y
OpenClaw:
Enabled github tool
Installed code_review skill
Try it: Review PR #42
Installing a skill may require enabling additional tools. Review those permissions before confirming.
Uninstall a Skill
Remove a skill when you no longer need it:
You: Uninstall skill blog_writer
OpenClaw: Uninstalled blog_writer skill
Creating Custom Skills
Skills are YAML files that define workflows. Create one when your process requires a repeatable sequence of tools and actions.
Skill Structure
name: daily_standup
description: Generates daily standup report
tools:
- jira
- github
- slack
steps:
- action: query_jira
query: "assignee = currentUser() AND status = 'In Progress'"
- action: query_github
query: "author:me is:pr is:open"
- action: format_report
template: |
Daily Standup Report
**Yesterday:**
{completed_tasks}
**Today:**
{in_progress_tasks}
**Blockers:**
{blockers}
- action: post_slack
channel: "#standup"
message: "{report}"
A custom skill typically defines:
- A name and description.
- The tools it is allowed to use.
- An ordered list of actions.
- Inputs, queries, and output templates.
Example: Custom Backup Skill
name: backup_project
description: Backs up project files to cloud storage
tools:
- zip
- exec
steps:
- action: create_zip
source: "."
output: "backup-{date}.zip"
exclude:
- node_modules
- .git
- action: upload
command: "aws s3 cp backup-{date}.zip s3://my-backups/"
- action: notify
message: "Backup complete: backup-{date}.zip"
Because this skill uses exec, review the command and enable the tool only if the workflow is trusted.
Install a Custom Skill
Save the YAML file and install it:
mkdir -p ~/.openclaw/skills
# Save the file as:
# ~/.openclaw/skills/backup_project.yaml
openclaw skill install ~/.openclaw/skills/backup_project.yaml
Run it with:
You: Run backup_project skill
OpenClaw: [Executes backup workflow]
Combining Skills
Combining skills lets you build larger automation workflows.
Example: Automated Code Review and Deployment
You: When a PR is merged to main, review the code, run tests, and deploy to staging
OpenClaw can create a workflow such as:
- Monitor GitHub for merges to
main. - Run the
code_reviewskill. - Run the
test_generatorskill. - If tests pass, run the
deployment_watcherskill. - Post the results to the
#engineeringSlack channel.
For API-focused projects, you can extend this workflow by integrating Apidog to run API contract tests and update documentation after successful deployments.
Example: Content Pipeline
You: Every Monday at 9 AM, research trending AI topics, write a blog post, and create a Twitter thread
The resulting workflow could:
- Schedule execution for Monday at 9 AM.
- Run a trend-monitoring skill.
- Run
blog_writer. - Run
social_media_manager. - Send the draft for approval.
Troubleshooting
Skill Fails to Install
Error:
Missing required tools
Fix: Enable the required tools, then install the skill again:
You: Enable tools: github, slack
OpenClaw: Enabled github and slack
You: Install skill code_review
OpenClaw: Installed code_review
Skill Does Not Work as Expected
Error:
Skill execution failed
Fix: Inspect the skill logs:
You: Show logs for code_review skill
OpenClaw: [Shows detailed execution logs]
Use the logs to identify the failing step and check whether the required tool has permission to perform that action.
Tool Permission Denied
Error:
Permission denied: exec tool
Fix: Enable the tool before running the skill:
You: Enable exec tool
OpenClaw: Enabled exec tool
Best Practices
1. Start with Safe Tools
Begin with read-only or low-risk tools:
readlistsearchweb_searchweb_fetchmemory
Add write and execution tools only when your workflow requires them.
2. Install Skills Gradually
Do not install all 53+ skills at once. Start with two or three skills that solve immediate problems.
3. Review Skill Definitions
Before installing a community skill, inspect its YAML file. Check:
- Which tools it requests.
- Whether it writes files or modifies external systems.
- Which commands it executes.
- Where it sends notifications or data.
4. Monitor Usage
Check which tools and skills are being used:
You: Show tool usage stats
OpenClaw:
Tool usage, last 7 days:
- read: 342 times
- web_search: 89 times
- write: 45 times
- exec: 12 times
Unexpected usage can indicate that a workflow needs tighter permissions or review.
5. Create Skill Aliases
Create shorter commands for frequently used workflows:
You: Create alias "review" for "run code_review skill"
OpenClaw: Alias created. Use "review" to run code_review.
FAQ
Can I disable all tools?
Yes, but OpenClaw will not be able to do much. It needs at least memory and context to function.
Are skills safe?
Community skills may be reviewed, but always inspect the code before installing one. A malicious skill could misuse the tools available to it.
Can I modify existing skills?
Yes. Copy the skill YAML to ~/.openclaw/skills/ and edit the copy.
How many skills can I install?
There is no stated limit, but too many skills can slow down OpenClaw. Install only the skills you use.
Do skills cost money?
Skills are free, but they may use tools that cost money. For example, image_gen may use API credits.
Can I share custom skills?
Yes. You can publish custom skills to the OpenClaw community repository on GitHub.
What if a skill breaks?
Uninstall it and report the issue to the skill author.
Can skills access my files?
Only if you enable the read or write tools. Skills cannot bypass tool permissions.
Conclusion
Tools give OpenClaw capabilities. Skills organize those capabilities into repeatable workflows. Together, they turn OpenClaw from a chatbot into an automation system.
A practical starting point is:
- Enable safe tools such as
read,search,web_search, andmemory. - Install one or two skills, such as
code_reviewormeeting_summary. - Review tool usage and skill logs.
- Enable higher-risk tools such as
execordatabaseonly when necessary. - Create custom YAML skills for recurring workflows.
The 25 tools and 53+ skills are a starting point for OpenClaw automation. As the community adds new capabilities, review each skill's permissions before adding it to your environment.
Top comments (0)