File Management and Code Execution Skills

File Operations

OpenClaw can interact with your local filesystem through four core tools: read, write, edit, and apply_patch.

Reading Files

Ask OpenClaw to read any file:

• "Read my notes.txt file"
• "What's in the config.json?"
• "Show me the last 20 lines of the log file"

The read tool is read-only and safe — it can't modify anything.

Writing Files

Create or overwrite files:

• "Create a file called ideas.md with my brainstorming notes"
• "Save this code to utils.py"
• "Write the meeting notes to meetings/2026-02-18.md"

Editing Files

Make targeted changes without rewriting the entire file:

• "Change the port number in config.json from 3000 to 8080"
• "Add a new entry to my grocery list"
• "Fix the typo in paragraph 3 of my essay"

Shell Commands

Execute terminal commands through natural language:

• "List all files in my Documents folder"
• "How much disk space do I have left?"
• "Find all Python files modified in the last week"
• "Run the test suite"

Safety

Shell commands can modify your system. OpenClaw's permission system requires confirmation for potentially destructive operations.

Code Execution

Run code snippets in a sandboxed environment:

• "Calculate the compound interest on $10,000 at 5% for 10 years"
• "Parse this CSV data and show me the averages"
• "Run this Python script and show me the output"

Practical Workflows

Project Organization

"Create a folder structure for a new web project with src, tests, docs, and config directories"

Batch Renaming

"Rename all .jpeg files in the photos folder to use a YYYY-MM-DD prefix based on their creation date"

Log Analysis

"Read the server.log file and summarize any errors from the last 24 hours"

Config Management

"Read my .env file and check if all required variables are set based on the .env.example"

Permissions

Control which file operations are allowed:

{
  "tools": {
    "read": { "enabled": true },
    "write": { "enabled": true, "confirm": true },
    "edit": { "enabled": true, "confirm": true },
    "shell": { "enabled": true, "confirm": true }
  }
}

Setting confirm: true means OpenClaw will ask before executing.