- Home
- Skills
- Agents & Orchestration
- Claudex: CEO Harness for Claude Code
Works with the AI Agents you already use
Claudex: CEO Harness for Claude Code
A CEO harness for Claude Code to route, delegate, and verify work across all your repositories from one session.
$15
Claudex: CEO Harness for Claude Code
Example session with this skill installed
/delegate my-api "Add a health check endpoint at /health that returns 200 OK"
- Read your context and instructions
- Compiled the claudex
I have routed this request to the my-api project and initialized a new delegation flow.
Task Delegation: T-1024
The task has been created in the ledger and a headless worker has been dispatched. I will monitor the process until git evidence is produced.
| Field | Value |
|---|---|
| Task ID | T-1024 |
| Project | my-api |
| Status | dispatched |
| Worker Type | Headless (claude -p) |
| Permission Mode | allowed_tools |
| Verification Gate | Git commit + npm test |
Directive sent to worker:
Implement a new GET
/healthendpoint. It must return a 200 OK status. Follow existing Express controller patterns insrc/routes/. Ensure a unit test is added intest/routes.test.jsand passes. Commit asfeat: add health check endpoint.
Next steps
- Wait for the worker to finish; I will notify you when the
reviewgate is ready. - Use
/sweepto check the status of this and other active background tasks. - View the full task requirements at
ledger/tasks/T-1024.md.
Connects securely to your tools. The creator never sees your data.
About this skill
Managing multiple repositories often leads to context switching, untracked agent runs, and unverified code changes. Claudex solves this by turning a single Claude Code session into a "CEO harness" that centralizes project management, routes work to specific repos, and delegates tasks to headless workers.
What it does
- Multi-project Routing routes requests to the correct repository based on a central registry.
- Headless Dispatch executes work in background
claude -psessions to keep your main terminal free. - Evidence-based Tracking closes tasks only when git commits, changelogs, or tests prove completion.
- Engineering & Marketing Doctrine applies specialized CTO/CMO playbooks to ensure high-quality code and safe content.
- Automated Reconciliation walks the ledger to find stalled tasks or unrecorded work via ground-truth git logs.
How it works
- Initialize the harness in a central directory to create your project registry and task ledger.
- Onboard existing git repositories to define their "trust dial" and allowed bash tools.
- Delegate work by creating a ledger task, which triggers a headless worker in the target repo.
- Verify results through automated review gates before the CEO session marks the task as done.
Frameworks & tools
This skill integrates with the claude CLI, git, bash, jq, and optionally Codex CLI for multi-model deliberation.
Why this beats prompting it yourself
Manual prompting across repos leads to "hallucinated progress" where agents claim tasks are done without committing code. Claudex enforces a hard split between the "deciding" CEO and "executing" worker, ensuring every action is recorded in a tamper-evident ledger.
Use cases
- Manage 10+ active repositories from a single unified command interface.
- Run multi-model "Fusion" panels to stress-test architectural decisions before implementation.
- Automatically generate project backlogs based on recurring defect classes and skipped tests.
- Gate autonomous marketing drafts to ensure brand safety and human approval.
Known limitations
Requires a local environment with bash and the claude CLI installed. Workers are limited by the allowed_tools defined in each project's dossier.
Requires local bash, git, and jq. Headless workers are constrained by project-specific allowed_tools. Does not support non-git repositories.
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
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 13 days ago
Requires