- Home
- Skills
- Agents & Orchestration
- Google Jules Orchestrator
Google Jules Orchestrator
A full-cycle development orchestrator that gates plans, reviews PR diffs, and iterates until code is merge-ready.
Free
Works with the AI Agents you already use
Google Jules Orchestrator
Example session with this skill installed
Run the Jules development loop to fix the retry logic in the sync service. The backoff counter isn't resetting on success. Follow the iteration cap of 3.
- Read your context and instructions
- Compiled the google jules orchestrator
I have initialized the orchestration loop for the sync service fix. The task is scoped, and the session is active.
Orchestration State: Task-SYNC-992
| Stage | Status | Details |
|---|---|---|
| Scoping | Completed | Defined fix for sync.py::push() retry loop. |
| Session | Created | Session ID: sess_88291-jules |
| Plan | Approved | Plan ID: plan_001 - includes test case for counter reset. |
| Implementation | In-Progress | Jules is currently modifying sync.py. |
| Review | Pending | Awaiting PR publication. |
Plan Summary
- Locate
push()function insync.py. - Add
self.backoff_counter = 0inside the successful request block. - Verify with a mock success after three failed attempts.
Next steps
- Poll
get_sessionto monitor implementation progress. - Once status is
COMPLETED, I will callpublish_prand begin the diff review. - Check
references/code-review-rubric.mdto ensure the fix doesn't introduce side effects in the sync loop.
Connects securely to your tools. The creator never sees your data.
About this skill
Development agents often drift from the original spec or submit messy pull requests that require manual fixing. This skill implements a strict orchestration loop that separates scoping and review from raw code implementation. It prevents implementation models from marking their own homework, ensuring every line of code is verified against a defined spec before merging.
What it does
- Task scoping creates a closed spec with explicit acceptance criteria to prevent scope creep.
- Session management handles the lifecycle of Jules implementation sessions via MCP tools.
- Plan gating enforces a mandatory approval step where the orchestrator reviews implementation plans before execution.
- Automated PR publishing triggers the native
publish_prflow once coding tasks are completed. - Loop-based iteration analyzes diffs against a rubric and generates structured fix-it prompts for re-implementation.
- Automated merging handles the final GitHub merge flow via CLI or MCP once code passes all checks.
How it works
- Scope the task using the provided template to define files, goals, and checkable criteria.
- Execute the Jules session by creating the task, approving the plan, and polling for completion.
- Audit the resulting pull request diff using a logic-dense code review rubric.
- Resolve issues by sending targeted fix-it prompts back to the session or merging the clean PR via GitHub.
Frameworks & tools
This skill requires the Jules MCP server for session control and the GitHub MCP server or gh CLI for repository management. It uses structured Markdown templates for state tracking and scoping.
Why this beats prompting it yourself
Manual prompting often leads to "hallucinated progress" where an agent claims success without verification. This skill forces a rigorous check-and-balance system that treats the implementation agent as a junior dev and the orchestrator as a senior reviewer.
Use cases
- Automating the end-to-end fix for reported GitHub issues or bug tickets.
- Implementing new features across specific modules while maintaining strict architectural boundaries.
- Refactoring legacy codebases where every change must be gated by a plan review.
- Managing high-volume repository maintenance tasks that require PR creation and merging.
Known limitations
Requires an authenticated Jules MCP server and GitHub access. Merging logic defaults to squash but follows repo conventions. Handover is mandatory if iteration caps are reached.
Requires Jules MCP and GitHub MCP/CLI. Iteration cap defaults to 4 attempts. Jules must be pre-registered with target repo sources.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 28 days ago
- Free with an Agensi account
Requires
Permissions
Allowed Hosts