Quorum: plan, build, review, adjudicate for Claude Code
Plan, build, multi-lens review, and adjudication - run step by step or autonomously from an approved plan, ending in a pull request.
$15.99
Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsQuorum: plan, build, review, adjudicate for Claude Code
Example session with this skill installed
/quorum:1-plan "Add a JWT-based authentication middleware to the Express API. It should check the Authorization header and attach the user object to the request."
- Read your context and instructions
- Compiled the quorum
I have investigated the request and prepared a work plan. I created the branch feat/jwt-auth-middleware and initialized the work directory.
Work Plan: feat/jwt-auth-middleware
| Field | Value |
|---|---|
| Slug | feat/jwt-auth-middleware |
| Status | planned |
| Author | claude-code |
Intent
Implement a robust JWT authentication middleware for the Express application to protect downstream routes and provide user context.
Acceptance Criteria
- Middleware extracts Bearer token from
Authorizationheader. - Returns
401 Unauthorizedif token is missing or malformed. - Returns
403 Forbiddenif JWT verification fails (expired/invalid). - Decoded payload is attached to
req.user. - Integration tests verify successful auth and all failure modes.
Proposed Design
graph TD
A[Incoming Request] --> B{Has Header?}
B -- No --> C[Return 401]
B -- Yes --> D{Verify JWT}
D -- Invalid --> E[Return 403]
D -- Valid --> F[Attach user to req]
F --> G[Next Middleware]
Next steps
- Review the plan above and the generated
docs/work/feat/jwt-auth-middleware/plan.md. - Approve the plan by running
/quorum:2-build feat/jwt-auth-middleware. - The build step will start implementing the middleware once you confirm.
Connects securely to your tools. The creator never sees your data.
About this skill
Development velocity often craters during the review and adjudication phase, where code intent gets lost in translation. Quorum is a complete pipeline for Claude Code that enforces a rigorous plan-build-review-adjudicate workflow to ensure production code matches user intent. It eliminates the "LGTM" culture by running every change through six independent, specialized review lenses—behavior, correctness, security, and more—before a final adjudication step resolves findings.
What it does
- Structured Planning turns requests into testable intent, acceptance criteria, and Mermaid diagrams before a single line of code is written.
- Autonomous Implementation builds features step-by-step from approved plans, tracking progress to maintain state across sessions.
- Multi-Lens Review spawns independent subagents to audit code through six specific remits, including behavioral testing and security.
- Automated Adjudication acts as a judge to verify review findings, apply accepted fixes, and generate a final verdict.
- Regression Guarding installs CI gates and mechanical checks to ensure requirements, tests, and review records remain untampered.
How it works
- Plan Use
/quorum:1-planto investigate the request and lock down acceptance criteria indocs/work/<slug>/plan.md. - Review Execute
/quorum:3-reviewto generate independent review files for the built change. - Adjudicate Run
/quorum:4-quorumto process review findings and finalize the code. - Monitor Use
/quorum:statusor/quorum:historyto track the state of work items across the repository.
Frameworks & tools
The pipeline uses Claude Code plugins, Python helpers, and shell scripts. It integrates with Git for branch management and supports standard CI/CD environments like GitHub Actions and GitLab CI.
Why this beats prompting it yourself
Standard prompts often suffer from self-confirmation bias, where the agent assumes its own code is correct. Quorum forces independence by spawning subagents with fresh context and specific audit remits, ensuring defects are caught before they reach a human reviewer.
Use cases
- Complex Refactoring managing multi-file changes where intent must be strictly preserved.
- Security-Sensitive Features requiring an independent security lens audit before merging.
- Long-Running Tasks needing persistent state and progress tracking across multiple Claude sessions.
- CI/CD Hardening automating the generation of behavioral regression tests and merge gates.
Known limitations
The pipeline requires installation as a Claude Code plugin to access ${CLAUDE_PLUGIN_ROOT}. All artifacts are stored locally in the repository under docs/work/.
Open source
Quorum is also available free and open source (MIT) at https://github.com/AiFirstDevelopment/quorum. This listing contains the same code with no exclusive functionality; you are paying for a packaged, ready-to-install copy.
What's included (14 skills in this plugin)
quorum-complete
Use when the user asks to install, set up, or learn about the Quorum plugins (quorum and tests) bundled in this folder - a plan, build, multi-lens review and adjudication pipeline for Claude Code, plus a regression-testing plugin. This skill only installs and explains the bundle; it does not run the pipeline itself.
1-plan
Step 1 of the quorum pipeline. Settles the work branch with the user - naming it, or starting a fresh one off the base when the current branch already carries a finished change - then investigates the request and writes intent, acceptance criteria, and non-goals to docs/work/<slug>/plan.md, with a mermaid diagram wherever one carries the design better than a paragraph. Writes no code.
2-build
Step 2 of the quorum pipeline. Implements the plan at docs/work/<slug>/plan.md, ticking off steps as it goes and recording any deviation from the plan.
3-review
Step 3 of the quorum pipeline. Reviews the built change from several independent lenses in fresh context, writing one review file per lens to docs/work/<slug>/reviews/. Finds problems; fixes nothing.
4-quorum
Step 4 of the quorum pipeline. Acts as judge over the plan, the code, and all reviews - adjudicating each finding, applying accepted fixes, and writing a verdict to docs/work/<slug>/verdict.md. Ends only with a green test suite.
audit
Measure an existing repository against a spec - free text, or a spec file already in the repo - and report where the implementation falls short, or "All clear". Runs on the default branch, writes no code, and works on a repo that never used the pipeline. Use to check that a repo faithfully implements a spec.
calibrate
Measure the review panel against fixtures whose defects are known in advance - catch rate and false-positive rate per lens. A deliberate, human-invoked evaluation, not part of any test suite. Use to decide whether the six-lens panel is worth its cost, or after changing a lens remit.
guard
Runs the mechanical checks on the current work item - requirements unchanged, no test weakened, reviews append-only, verdict self-consistent, evidence real, enforcement posture stated - and can install them as a CI gate. Reports violations; fixes nothing.
history
Lists every change this repository has planned through the quorum pipeline, oldest first - what it was, who planned it, when, and where it got to. Reads docs/work/ and git history. Read-only; changes nothing.
pipeline
Runs the whole quorum pipeline autonomously from an approved plan - build, six independent review lenses in parallel, adjudication, then a read-only recheck of the judge's own commits - stopping only for plan approval at the start. Use when the plan is written and you want the change delivered without further supervision.
status
Reports where the current work item stands in the quorum pipeline - what has run, what it concluded, and what has changed since - and names the next command to run. Read-only; changes nothing.
add
Writes behavioral regression tests for the current change, exercising the fully assembled application through its outermost user-facing surface. Verifies each new test actually fails when the behavior it guards is broken. Use when asked to add tests, harden a change, or close test gaps.
ci
Wires the regression suite into CI as a required status check so a pull request cannot merge while tests fail. Generates the workflow file and explains the branch-protection settings that must be applied in the hosting UI or API.
run
Runs the full regression suite, discovering and recording how this repo runs its tests, and reports failures in a structured form the rest of the quorum pipeline can consume. Distinguishes broken code from broken tests from flakes.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scanned
Verified clean 1 day ago
- Passed all security checks, Safe to install
Requires
Python
Git