| name | jules-cli |
| description | Interface with the Google Jules CLI to delegate asynchronous coding tasks. Use this tool to execute a specific plan, start a background session, or retrieve results from a remote Jules agent. |
Jules CLI Skill
Goal
To execute the jules command-line tool for delegating long-running or complex coding tasks to a remote agent session. This skill handles session creation, status monitoring, and result retrieval.
Usage Rules
1. Starting a Task
- Command:
jules remote new
- Required Flags:
--session "<prompt>": The detailed instruction or plan you want Jules to execute.
--repo .: Specifies the current directory as the target repository.
- Constraints: Always quote the session prompt. If the prompt is complex (multi-step), ensure it is passed as a single string.
2. Non-Interactive Mode
- NEVER run
jules without arguments (this opens a TUI).
- ALWAYS use the subcommand syntax
jules remote <action>.
3. Monitoring & Retrieval
- Use
jules remote list --session to check the status of running tasks.
- Use
jules remote pull --session <ID> to apply the changes (PR/Branch) locally once the task is complete.
Command Reference
| Action | Command Pattern |
|---|
| Start New Session | jules remote new --repo . --session "<Your Plan>" |
| List Sessions | jules remote list --session |
| Pull Results | jules remote pull --session <Session_ID> |
| Check Version | jules version |
Examples
Example 1: Execute a Specific Plan
User: "Here is the refactoring plan for the auth module. Run it with Jules."
Agent Execution:
jules remote new --repo . --session "Refactor src/auth.ts to use the Singleton pattern. Update all import references in src/routes/ to match the new class structure. Run tests to verify."