AgenKitkits

Kit · v2.2.0

the software team

Engineering Kit

A guided, specialist-driven engineering workflow: brainstorm to architecture to plan to test-first build to review to ship, run by ten deep specialist agents.

Claude CodeCursorCodex CLIOpenCodeAntigravity CLIGitHub CopilotKimi Code

The workflow

What happens when you run /agenkit.

You describe the task once. The Conductor runs the whole engineering lifecycle, handing each phase to the specialist who owns it and stopping for your approval before it moves on.

One command, six phases

Nothing is written until you approve the spec, the architecture, and the plan. Every task is built test-first and reviewed twice: once for spec compliance, once for code quality.

Build any software, at any scale - solo.

In Claude Code · Cursor · Codex · +4 more →
/agenkit
  1. OrientContext Engineer
    no memory yet, indexing your codebase
  2. BrainstormConductor
    turns your idea into a clear spec
  3. ArchitectureBackend Architect
    services · data model · idempotency
  4. Build · test-firstTest Engineer
    failing test → green → refactor
  5. ReviewCode Reviewer
    spec ✓ then code quality ✓
  6. ShipInfra / DevOps
    CI/CD · deploy · verify
Running the pipeline…

The six phases, in detail

  1. 01
    BrainstormConductor

    Asks focused questions, one at a time, until your idea is a written spec: scope, constraints, and what success means. No code is touched here.

    You approve the spec
  2. 02
    ArchitectureBackend Architect

    Designs the folder structure, module boundaries, interfaces, and data model, plus the failure strategy: idempotency, reconciliation, retries, timeouts.

    You approve the architecture
  3. 03
    PlanConductor

    Turns the approved spec and architecture into bite-sized, test-first tasks with exact file paths. Nothing vague survives this step.

    You approve the plan
  4. 04
    BuildTest Engineer + domain specialists

    Every task runs red, green, refactor: a failing test first, the minimal code to pass it, then cleanup. Work is committed as it goes.

    Tests pass on every task
  5. 05
    ReviewCode Reviewer

    Two verdicts on every change: does it match the spec, and is it well built. Critical and important findings block until they are fixed.

    Review must pass
  6. 06
    ShipInfra / DevOps Engineer

    CI/CD, deployment with verification and rollback, and a final security and performance check before anything merges.

    You choose merge or PR

How to use it

From install to shipped.

  1. 01
    Subscribe and activate

    Checkout issues a license key. Run agenkit activate <key> once; it's stored locally. There is no account to manage.

  2. 02
    Install into your project

    Run agenkit install engineering-kit in your project folder. It auto-detects your harness and writes files where that tool expects them.

  3. 03
    Open your harness

    Claude Code, Cursor, Codex CLI, OpenCode, Antigravity, GitHub Copilot, or Kimi Code. The same team shows up in all seven.

  4. 04
    Run /agenkit

    Type /agenkit and describe what you're building in one sentence (in GitHub Copilot use @agenkit; in Codex, ask for the AgenKit skill). That's the only thing you have to define.

  5. 05
    Approve at each checkpoint

    You sign off on the spec, the architecture, and the plan before a single line is written. You stay in control the whole way.

  6. 06
    Ship, then keep current

    Merge or open a PR once review passes. Run agenkit update later to pull new versions.

Every specialist you can call by name

Let the Conductor route the work, or summon one directly: “ask the backend-architect to review this webhook handler”.

backend-architectcode-reviewerconductorcontext-engineerdata-engineerdebuggerfrontend-engineerinfra-devops-engineerperf-observability-engineersecurity-engineertest-engineer

Who it's for

Built for the people shipping alone.

One subscription, the same specialist team in every harness you work in.

◆ Indie hackers & solo founders

Shipping a real product alone. The kit is the engineering team you haven't hired: architecture, tests, review, and deploy on every feature.

⌘ Freelancers & agencies

Client work that has to hold up in production. The gated workflow turns one-person delivery into a reviewable, documented process.

≈ Engineers with a day job

Side projects and internal tools built with senior-team discipline: test-first, properly reviewed, without waiting on anyone.

The kit encodes the working discipline of senior engineering teams: the test-first loops, two-stage code review, architecture gates, threat models, and SLO thinking you'd expect at a top tech company or a quant desk. And it doesn't sit still: releases are versioned and continuously improved, and every update is included while you're subscribed. Run agenkit update and the whole team gets sharper.

Compatibility

Installs into every harness.

HarnessInstall destination
Claude Code.claude/{agents,skills,commands}
Cursor.cursor/commands + .claude/{agents,skills}
Codex CLI.agents/skills + .codex/agents
OpenCode.opencode/{agents,commands,skills}
Antigravity CLI.agents/plugins/engineering
GitHub Copilot.github/{agents,skills}
Kimi Code.agenkit/<kit> · /plugins install

Installation docs · /agenkit workflow · Specialist agents · Guides