- Home
- Skills
- Agents & Orchestration
- Agent Team Lead
Agent Team Lead
Turn your AI into the lead of a small agent team: split work, run Codex and Claude Code in parallel, review across vendors, and verify before accepting.
Free
Works with the AI Agents you already use
Agent Team Lead
Example session with this skill installed
Add CSV export to this project. Split the work between Codex and Claude Code, run what you can in parallel, and have the other vendor review the code.
- Read your context and instructions
- Compiled the agent team lead
Summary
Example run, shortened. Two tasks ran in parallel, a dependent task followed, and an agent from another vendor reviewed the work.
$ python3 scripts/team.py run plan.json
Run 20261106-001311 started from 7d30691d27
[running] export-core
[running] export-docs
[passed] export-docs
[passed] export-core
[running] review-export
[passed] review-export
[final verify] passed: python3 -m unittest
| Task | Agent | Result |
|---|---|---|
| Write the CSV export function | Codex | Tests passed |
| Document the export feature | Claude Code | Passed |
| Review the export code | Claude Code | Report ready, read-only |
The lead then checks the work instead of trusting the reports:
git diff --stat 7d30691d27 agent-team/20261106-001311/integration
python3 -m unittest
If an agent edits a file outside its allowed list, the attempt is rejected. If a verify command fails, the failure output goes back to the agent for one retry.
Next steps
- Read the review report and check each finding
- Merge with
git merge --squashonce you are satisfied - Clean up with
python3 scripts/team.py clean <run_id>
Connects securely to your tools. The creator never sees your data.
About this skill
A lightweight, portable way to make several coding agents work like a team, with one AI acting as the lead.
The lead breaks down the work, gives each task to the right agent, runs independent tasks in parallel, has a different vendor's agent review the result, and runs the checks itself before accepting.
What's included:
- A task runner that needs only Python and git, with nothing else to install
- Built-in support for Codex and Claude Code, and a simple config to add other agent CLIs
- An isolated git worktree for each task, so parallel agents do not overwrite each other
- Checks that reject work outside a task's allowed files, plus automatic retry with the failure output
- Falls back to your host's subagents, or to working alone, if no other agent CLI is installed
Your current branch is never changed. Finished work waits on a separate branch until you merge it.
What it runs: the script starts the agent CLIs and verify commands named in your plan, on your machine. It makes no network calls itself. Agent CLIs use your own accounts and credits. It works on macOS, Linux and WSL.
This is the lite edition of xay, an open-source multi-agent workflow: https://github.com/Flame929/xay-multiagent
Runs on macOS, Linux or WSL, not native Windows. Writing tasks need a git repository with at least one commit. Agents start from the last commit, so uncommitted changes are not visible to them. Only Codex and Claude Code are built in, and other agent CLIs need a small config. Each agent uses your own subscription or API credits.
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
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 11 days ago
- Free with an Agensi account
Requires
Permissions
The script starts the agent CLIs and verify commands named in your plan. Agents edit files in isolated git worktrees. The script makes no network calls itself, but agent CLIs use your own accounts and network access.