Agent Infra CLI
Shared infrastructure for Claude Code, Codex, Antigravity CLI and OpenCode: a task lifecycle, skills and sandboxes.
- 87 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Skills create tasks, branches and pull requests in your repository
- Sandboxes and the Feishu command bridge route requests to service processes
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/agent-infraDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Skills create tasks, branches and pull requests in your repository
- Sandboxes and the Feishu command bridge route requests to service processes
This is third-party code. Review the repository files before installing.
What it does
Agent Infra gives several AI agents in one repository the same commands, skill vocabulary, project governance files, isolated sandboxes and upgrade path. `ai init` prepares a project once, then skills drive the work: from an issue in the tracker to a merged pull request along a path the analysis selects by risk, with independent code review in every path. The lifecycle is built from `/import-issue`, `/analyze-task`, `/plan-task`, `/code-task`, `/review-code`, `/create-pr` and `/complete-task`. It works the same in Claude Code, Codex, Antigravity CLI and OpenCode, and task state follows the agent. There is a command bridge for Feishu, bilingual documentation and an archive of process data from local tasks and GitHub evidence.
Who it is for. Teams that mix several AI agents in one repository and want a shared way of working and an auditable task history.
Good fit when
- A tracker issue goes from analysis to a merged pull request with independent code review.
- Developers work in different agents while the process must stay the same.
- A project gets development sandboxes and controlled skill runs.
- After a template update the latest skills should be pulled with `/update-agent-infra`.
Not a fit when
- A single developer works with a single agent: the shared layer adds needless procedure.
- Installation without Node.js 22.9 or newer is required.
- The task is one-off and needs no trail in issues and pull requests.
Example request
Import issue 42 from the tracker, analyze it, select a work path and show the plan before any edits.Limitations
Needs Node.js 22.9.0 or newer. Install with `npm install -g @fitlab-ai/agent-infra`, the `install.sh` script or Homebrew: `brew install fitlab-ai/tap/agent-infra` after `brew trust fitlab-ai/tap`. The demo image is stored in Git LFS and needs `git lfs pull` to view. `ai update` updates the tool itself, and `ai sync` refreshes project files. The process is strict: every task goes through analysis and independent review, which adds steps even to small edits.
How to disable. Remove the tool with `npm uninstall -g @fitlab-ai/agent-infra` and delete `.agents/` files and skills in the project by hand if they are no longer needed.
FAQ
How do I install it?
With `npm install -g @fitlab-ai/agent-infra`. There is also an `install.sh` script and the Homebrew formula `fitlab-ai/tap/agent-infra`.
Where do I start in a project?
Run `ai init` at the root: the tool collects project details and prepares files. Then run `/update-agent-infra` in the agent to pull the latest skills.
Which work paths exist?
Analysis selects a path by risk, from short to full `analyze, review-analysis, plan, review-plan, code, review-code`. Independent code review is part of every path.
How do I update it?
`ai update` updates the tool itself, and in an initialized project `ai sync` refreshes template files and the registry.
Which agents are supported?
Claude Code, Codex, Antigravity CLI and OpenCode. Task state carries over when you switch tools mid-task.
Discussion
No questions or reviews yet.
Related
CLI toolsA plugin for Claude Code and Codex: a shared task wall, project memory, reports and a dashboard that shows agents' work across sessions
DevChain runs a team of AI coding agents on your own machine with a shared board, chat, code review and live terminals.
A bridge between local CLI agents: Codex, Cursor or Claude Code delegate work to Antigravity, Grok, Kimi and seven more.
An open coding agent for the terminal, browser and desktop that works with models from OpenAI and Claude to Ollama
A local control panel for coding agents: it runs Claude Code and Codex, keeps a task queue and merges verified work into main.
AI Builder Club Skills
AI Builder Club — Skills
A Claude Code plugin marketplace to make a repo agent-ready and run compounding agent loops
README in short
Covers why it exists, an example task, key features, the sandbox control boundary, a quick start and updating.