By petekp
Structured, resumable, multi-stage developer flows for Claude Code. /circuit:run selects the best flow for a natural-language task and invokes /circuit:explore, /circuit:review, /circuit:migrate, /circuit:fix, /circuit:build, or /circuit:sweep; /circuit:create and /circuit:handoff cover custom flows and continuity.
Runs the Build flow directly through the project CLI, with optional Lite, Deep, or Autonomous entry behavior.
Drafts, validates, and optionally publishes a user-global custom Circuit flow.
Investigate, understand, choose among options, or shape an execution plan. Invokes the circuit-next `explore` flow end-to-end via the project CLI, producing a run trace + final report under the run folder.
Runs the Fix flow directly through the project CLI, with optional Lite, Default (standard), Deep, or Autonomous entry behavior.
Saves, resumes, clears, briefs, or installs hooks for Circuit continuity through the project CLI.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Structured flows for coding agents.
Circuit is an orchestration layer for structured, resumable, multi-stage
developer flows. Ask Circuit to handle a task in natural language. In
Claude Code, /circuit:run selects the right flow before it starts. In
Codex, @Circuit lets Codex choose the best bundled Circuit skill. In the
local CLI, circuit-next run --goal keeps the deterministic router path.
Once a flow is selected, Circuit runs the same headless engine in every host and checks each step's output against a contract before moving on.
Circuit is currently a pre-release alpha. Install it directly from a checkout:
git clone https://github.com/petekp/circuit-next.git
cd circuit-next
npm install
npm run build
Then run a flow through the local launcher:
./bin/circuit-next run --goal '<your task>'
The CLI router classifies your task and runs the right flow. To use the slash commands inside Claude Code, load the self-contained plugin package:
claude --plugin-dir ./plugins/claude
To use Circuit from Codex, install or refresh the Codex plugin package and ask
Codex to use @Circuit. Codex can choose the best bundled Circuit flow
skill from your natural-language request.
Optional but recommended: drop a personal config at
~/.config/circuit-next/config.yaml to set defaults (model, reasoning effort,
local skills, connector routing) across every project. A repo-local
./.circuit/config.yaml overrides those defaults per project. See
User-Space Configuration for details.
Circuit replaces ad-hoc skill invocation and copy-pasted instructions. Use one natural-language front door for normal work, or call a flow directly when you already know what you want.
Core Flows:
These flows ship with the plugin. Build, Fix, Explore, Review, Migrate, and Sweep can all be selected by the host model or invoked explicitly.
| Flow | Purpose |
|---|---|
| Explore | Investigate, understand, choose among options, shape a plan |
| Build | Features, refactors, docs, tests, mixed changes |
| Fix | Bugs, regressions, flaky behavior |
| Review | Audit-only review, no implementation |
| Migrate | Framework swaps, dependency replacements, architecture transitions |
| Sweep | Cleanup, quality passes, coverage improvements |
Modes:
Each flow accepts a mode at run time that scales how thorough the run is and which review steps fire.
| Mode | Behavior |
|---|---|
| Default | Plan, act, verify, independent review where the flow includes one. |
| Lite | Skips the review relay where the flow allows it. Use for small, low-risk changes. |
| Deep | More thorough analysis and review. Pauses for input at architecture-class checkpoints. |
| Tournament | Competing proposals with adversarial evaluation. Available on Explore and Review. |
| Autonomous | Checkpoints auto-resolve to safe defaults. Useful for unattended runs. |
Pass a mode with --entry-mode <lite|deep|autonomous> (or --mode, the
shorter alias). Tournament is a depth-level option exposed through
--depth tournament on the flows that support it. Mode availability varies
by flow; see each flow's src/flows/<id>/schematic.json for the
authoritative list.
Every flow is built from a fixed set of stages: Frame, Analyze, Plan, Act, Verify, Review, Close. Not every flow runs every stage, but the order holds.
The flow is selected. In host plugins, the host model may choose a flow before calling Circuit. In CLI router mode, Circuit's deterministic router chooses.
Steps run in the right order. Research before decisions. Decisions before implementation. Implementation gets an independent review from a separate worker. Every step writes a typed report and an entry in the run trace.
Progress survives session crashes. Each run gets its own folder with a trace, reports, and evidence. If a session dies, resume against that folder and Circuit picks up at the last completed step.
Stay in the loop. Flows pause at checkpoints when they need scope confirmation or a tradeoff decision. Everything else runs autonomously. Autonomous mode resolves checkpoints to their safe default and keeps going.
Default front doors:
npx claudepluginhub petekp/circuit-next --plugin circuitCompile a plain-language task into a concise, auditable Codex or Claude Code /goal with a clear definition of done, transcript-visible proof, constraints, bounded stop conditions, and risk-based review depth
Multi-phase workflows for Claude Code that survive session crashes, enforce review gates, and resume exactly where they stopped.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Develop, test, build, and deploy Godot 4.x games with Claude Code. Includes GdUnit4 testing, web/desktop exports, CI/CD pipelines, and deployment to Vercel/GitHub Pages/itch.io.
Consult multiple AI coding agents (Gemini, OpenAI, Grok, Perplexity, plus codex and gemini CLIs when installed) to get diverse perspectives on coding problems
Advanced development superpowers for orchestrating complex workflows with Superpower Loop and work verification
Complete collection of battle-tested Claude Code configs from an Anthropic hackathon winner - agents, skills, hooks, and rules evolved over 10+ months of intensive daily use