iterative-development plugin
A Claude Code plugin for large specs: requirements with proof obligations, a walking skeleton and audited sprints
- 183 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The autonomous loop runs subagents for hours that write code and commit to the repository; a human steps in only at iteration boundaries.
Install
- Works in
Manual install
/plugin marketplace add prime-radiant-inc/prime-radiant-marketplaceThen `/plugin install iterative-development@prime-radiant-marketplace` and a Claude Code restart.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The autonomous loop runs subagents for hours that write code and commit to the repository; a human steps in only at iteration boundaries.
This is third-party code. Review the repository files before installing.
What it does
The iterative-development plugin leads Claude Code through building a project from a large or ambiguous spec, where an ordinary plan loses the thread. The `extracting-requirements` skill splits the spec into chunks, extracts stories with proof obligations and behavior scenarios using parallel subagents, and builds a coverage ledger. The `scoping-the-simplest-core` skill defines a walking skeleton (iteration ITER-0000) and a roadmap. The skeleton has to close at least one end-to-end scenario, and compiling alone is not enough. Then the loop repeats. `running-an-iteration` runs a sprint from a sentinel baseline and scope review to implementation and final runs, `implementing-tasks` carries out each task with evidence, and `auditing-progress` checks the result with paired auditors on three levels. The loop ends when the roadmap is empty and a final audit confirms that every observable requirement has passing behavior evidence.
Who it is for. Developers and tech leads who hand Claude Code the job of building a product from a bulky spec and want the result checked against behavior evidence.
Good fit when
- The spec spans more than ten files and over a hundred requirements, and a single plan does not hold.
- The product is wanted in a working, testable state at every iteration boundary.
- The spec changes along the way: an edit is picked up at an iteration boundary while story IDs are preserved.
- The work is handed to an agent for several hours unattended and progress is followed in the iteration log.
- After a session crash work resumes with the phrase "continue iterative development with the existing plan", because the state lives in files.
Not a fit when
- The project is small and bounded: the `superpowers:writing-plans` and `superpowers:subagent-driven-development` pair is simpler for it.
- There is no spec: the plugin extracts requirements from existing material and does not invent the product itself.
- You want a result in minutes: a loop with subagents and paired auditors runs for hours.
- The token budget is limited: requirement extraction and audits launch several subagents.
Example request
Use iterative-development: the spec is in the `specs/` folder, extract the requirements, define the walking skeleton and start the first iteration.Limitations
The plugin is early: version 0.1.0, with the last commit made in June 2026. All state lives in files under `docs/superpowers/iterations/`, so the project gains a requirements backlog, scenarios, a roadmap and an iteration log. A human intervenes only between iterations: a change entered in the middle of a sprint is applied after it finishes. The method is designed to pair with the superpowers set (brainstorming, TDD, verification before completion), so it is worth installing alongside. The plugin's own checks run with `./scripts/run_validation_suite.sh`.
How to disable. Uninstall the plugin through the `/plugin` menu in Claude Code and restart it. The `docs/superpowers/iterations/` directory with project artifacts is deleted by hand if it is no longer needed.
FAQ
How do I install iterative-development?
In Claude Code add the marketplace with `/plugin marketplace add prime-radiant-inc/prime-radiant-marketplace`, then install the plugin: `/plugin install iterative-development@prime-radiant-marketplace`. Restart Claude Code afterwards.
When do I pick iterative-development over superpowers:writing-plans?
This plugin fits large, detailed or ambiguous specs where a single plan has failed before. For small bounded projects the planning and subagent-driven development pair from superpowers is simpler.
Where is the state stored and how do I resume after a crash?
In files under `docs/superpowers/iterations/`: the requirements backlog, scenarios, the `roadmap.md` roadmap and the `iteration-log.md` log. Just ask to "continue iterative development with the existing plan" and the loop picks the next pending iteration.
How do I change the spec in the middle of the work?
Type in the chat what changed, for example "the spec changed, re-read specs/foo.md". The orchestrator picks it up at the next iteration boundary, re-extracts the affected stories and revises the roadmap if needed. Removed stories are marked deferred rather than deleted.
Which skills does the plugin consist of?
Six: `iterative-development` as the orchestrator, `extracting-requirements`, `scoping-the-simplest-core`, `running-an-iteration`, `implementing-tasks` and `auditing-progress`.
Discussion
No questions or reviews yet.
Related
PluginsA Claude Code plugin: spec-driven development over a single SPEC.md file, a loop of seven commands with no sub-agents
A plugin with three skills for large development work: project analysis, phased tasks tracked as GitHub Issues, parallel agents and code review
A skill that interviews you and writes requirements.md, design.md and tasks.md so Claude Code and Cursor work from one plan
A Claude Code plugin that turns vague requirements into numbered spec files with a dependency manifest through an interview.
Claude Code plugins with development recipes: agree on the outcome, design, implement with checks and run an independent review
Nick Tune's plugin marketplace for Claude Code: checks before a task is handed in, code review and TDD personas
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| iterative-development (this card) | no | 183 | |
| Superpowers | no | 313,096 | a set of process skills: plan, TDD, subagents and review; iterative-development adds an audit loop for large specs |
| Spec Kit | no | 148,171 | GitHub's CLI for spec-driven development with commands and skills for the agent; here parallel subagents break down the spec and sprints go through an evidence audit |
| Ralph | no | 22,121 | a bash loop restarts the agent with clean context until the PRD stories pass checks; here state lives in artifacts and paired auditors verify the result |
README in short
The project page briefly names the use cases and the five stages of the loop, then lists the six skills with their roles. Separate sections explain installation through the Prime Radiant marketplace and the relation to superpowers.