Claude Code SDLC Harness plugin
Hooks and skills make Claude Code plan before coding, test before shipping and escalate when it is not confident.
- 53 ★ on the repository
- Last commit
Extended access
Why this level
- Hooks and scripts run as shell commands on each agent interaction.
- The wizard writes documents, skills and hook settings into the project.
Install
- Works in
Manual install
npx -y agentic-sdlc-wizard@latest initRun in the project root and restart Claude Code. The plugin installs with `claude plugin install sdlc-wizard@sdlc-wizard-marketplace`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Hooks and scripts run as shell commands on each agent interaction.
- The wizard writes documents, skills and hook settings into the project.
This is third-party code. Review the repository files before installing.
What it does
The set brings development discipline to Claude Code: hooks remind it about tests before writing code and skills walk a task through an SDLC checklist. Installation runs a setup wizard. It reads the project and creates `CLAUDE.md`, `SDLC.md`, `TESTING.md` and `ARCHITECTURE.md` for it, plus a `.claude/` directory with hooks, skills and settings. The plugin has six skills: `sdlc` drives development, fixes and releases, `setup` runs setup, `update` shows a changelog and applies changes selectively, `feedback` sends feedback, and `conduct` (aliased as `fleet`) coordinates Claude and Codex sessions for reviews and handoffs. For high-stakes changes the set requires independent cross-model review through Codex. The `agentic-sdlc-wizard check` command reports which files match the reference, were customized, are missing or drifted, and exits with an error for CI.
Who it is for. Teams writing code with Claude Code who want one consistent workflow: plan, tests, self-review, release.
Good fit when
- An agent started writing code without tests and the "tests first, implementation second" order must return.
- A new repository needs `CLAUDE.md`, `TESTING.md` and hooks created by one wizard.
- Before a high-stakes change ships you want an independent check by a second model.
- CI must confirm the agent setup files have not drifted from the reference.
Not a fit when
- You want a light rule set: hooks fire on every interaction and add tokens.
- The work does not run in Claude Code: the author keeps separate projects for Codex and OpenCode.
- The project is not ready for tests and review: the set insists on them.
Example request
Set up an SDLC process in this repository: create the documents and hooks, then take a task from the list and run it through the `sdlc` skill checklist.Limitations
The set targets Claude Code and installs with `npx -y agentic-sdlc-wizard@latest init`; the plugin installs with a separate command. The README notes a known defect: in a new project setup does not auto-run and the skill must be called manually. The recommended model combination is described as of November 2026 and may change. The repository states no license. Hooks and generated scripts run in your environment, so read them before installing. E2E scoring runs in the repository CI and does not land in your project.
How to disable. Remove the `sdlc-wizard` plugin through `/plugin`. Generated hooks, skills and documents live in `.claude/` and the project root: remove them and the hook entries in `.claude/settings.json` by hand.
FAQ
How do I install the SDLC Harness?
Run `npx -y agentic-sdlc-wizard@latest init` in the project root and restart Claude Code. A no-npm option: `claude plugin install sdlc-wizard@sdlc-wizard-marketplace`.
Why did setup not start in a new project?
It is a known defect: auto-invoke does not fire in a brand-new project. Type `/` plus the setup skill name to run it manually.
What does the setup wizard create?
`CLAUDE.md`, `SDLC.md`, `TESTING.md` and `ARCHITECTURE.md`, plus hooks in `.claude/hooks/`, skills in `.claude/skills/` and `.claude/settings.json`.
How do I check that setup has not drifted from the reference?
Run `npx agentic-sdlc-wizard check`, add `--json` for CI. For each file it reports MATCH, CUSTOMIZED, MISSING or DRIFT and exits non-zero on MISSING and DRIFT.
Is Codex required?
For high-stakes changes the set requires cross-model review through Codex. For other tasks it is optional. The author keeps a `codex-sdlc-wizard` project for Codex itself.
Discussion
No questions or reviews yet.
Related
PluginsA plugin that turns Claude Code into a 16 agent team: documentation first planning, TDD and 9 pre-merge quality gates.
Ten software lifecycle skills for Claude Code: requirements, design, tests, adversarial review and release checks with hooks.
The AI-SDLC plugin for Claude Code carries a task from spec to PR: a readiness gate, a developer agent and several reviewers.
A Claude Code plugin with /readiness and /setup skills that scores a repo for agent work and installs rules, hooks and checks.
ADLC Harness takes a task from intent to PR with one command, and the human approves only four key gates.
A handbook and starter set for Claude Code: guides, CLAUDE.md templates, starter kits, hooks and the commit-helper plugin.
README in short
The README presents the set as a self-improving process control system for AI agents. It walks through the layers: philosophy, hooks, scoring, statistical validation and self-update. Two model lanes, cross-model review and comparisons with the author's sibling projects are covered separately.