AI-SDLC plugin
The AI-SDLC plugin for Claude Code carries a task from spec to PR: a readiness gate, a developer agent and several reviewers.
- 441 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- The pipeline writes code, creates branches and commits, runs agents and opens pull requests by itself.
- Plugin hooks block some actions, but the agent operates on repository worktrees.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/ai-sdlcDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The pipeline writes code, creates branches and commits, runs agents and opens pull requests by itself.
- Plugin hooks block some actions, but the agent operates on repository worktrees.
This is third-party code. Review the repository files before installing.
What it does
AI-SDLC is a decision engine for spec-driven development. The Claude Code plugin adds subagents, slash commands and hooks. The `/ai-sdlc execute` command creates a git worktree, runs a developer agent and three reviewers and opens a pull request. Before dispatch a Definition of Ready gate refuses tasks the operator has not yet decided key questions for. The plugin ships `developer`, `code-reviewer` and `security-reviewer` subagents plus a review variant through Codex CLI for independence from the author. Commands cover conflict resolution, rebasing an open PR, issue triage, pipeline status, spec import and RFCs. The `cli-hierarchy` session hierarchy starts a planner and up to five executors, each in its own tmux Claude Code session. Configuration is declared as resources with JSON schemas.
Who it is for. Teams handing many backlog tasks to Claude Code in parallel who want to steward decisions and reviews rather than type code.
Good fit when
- A backlog task should reach an open PR without manual steps.
- An independent review by a different agent is needed, for example through Codex CLI.
- Up to five tasks run in parallel and should be steered from one place.
- A spec must become tasks with verifiable criteria.
Not a fit when
- The project is small with few tasks: configuring policies and gates will not pay off.
- There is no backlog and spec process for the agent to work from.
- Review and merging must stay fully manual with no automated agents.
Example request
Run `/ai-sdlc execute AISDLC-42` for a task from the backlog.Limitations
The project spec is at v1alpha1, so interfaces may change. The plugin works only in Claude Code, and plugin subagents cannot spawn other subagents, so `/ai-sdlc execute` creates the developer and reviewers from the main session. Review through Codex needs Codex CLI installed. The pipeline creates branches, commits and pull requests: connect it to a test repository first. Full runner and session hierarchy setup is covered in the site documentation.
How to disable. Remove the plugin with /plugin and the ai-sdlc marketplace. Delete the `.ai-sdlc` configuration files and repository hooks by hand.
FAQ
How do I install the AI-SDLC plugin?
In Claude Code run `/plugin marketplace add ai-sdlc-framework/ai-sdlc` and `/plugin install ai-sdlc@ai-sdlc`, then `/reload-plugins`. After that run `ai-sdlc init` in the repository.
What does /ai-sdlc execute do?
It creates a worktree, runs a developer, three reviewers and opens a PR. If a task fails the readiness check the dispatch is refused.
Which agents are supported?
The framework is agent-agnostic: Claude Code, Codex, Cursor, Copilot, Aider and any OpenAI-compatible API. The plugin with commands targets Claude Code.
What is the Definition of Ready?
A pre-dispatch check: a task does not go into work until the operator has decided what must not be left to the agent mid-execution.
How do the reviewers differ?
There are reviews for code quality, tests and security. Security runs on a stronger model by default, and the Codex variants give independence from the author.
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.
Hooks and skills make Claude Code plan before coding, test before shipping and escalate when it is not confident.
Claude Code project template: spec, design, code and deploy live in the repo, and ten skills guide the agent by phase.
Nine plugins for Claude Code with 51 skills and 2 agents: a single workflow from a rough idea to a reviewed pull request.
Ten software lifecycle skills for Claude Code: requirements, design, tests, adversarial review and release checks with hooks.
Claude Code plugin with 17 skills: task sizing, design review, TDD, fresh-context subagents and a PR after three human decisions.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| AI-SDLC (this card) | no | 441 | |
| Spec Kit | no | 149,633 | covers the path from idea to contract; AI-SDLC covers the path from contract to merged PR |
| Claude Squad | no | 8,790 | manages parallel agent sessions; AI-SDLC adds a readiness gate, review and attestation |
README in short
The project holds that decisions made upfront by an operator are cheaper and more accurate than AI decisions mid-execution. The framework has five pillars: the decision engine, session hierarchy, cross-harness review, the operator surface and declarative governance.