Help us improve
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
By lvlup-sw
Orchestrate a durable, stateful SDLC workflow for coding agents with convergence gates, agent teams, and full audit trail — from design exploration and TDD-based implementation to multi-stage code review, PR merge, and cross-session state checkpointing.
npx claudepluginhub lvlup-sw/exarchosStart collaborative design exploration for a feature or problem
Author an architectural invariant catalog entry through a guided interview
Run a lightweight oneshot workflow — plan + TDD implement + optional PR
Create TDD implementation plan from design document
Prune stale workflows from the pipeline (dry-run → confirm → apply)
Use this agent when dispatching TDD implementation tasks to a subagent in an isolated worktree. <example> Context: Orchestrator is dispatching a task from an implementation plan user: "Implement the agent spec handler (task-003)" assistant: "I'll dispatch the exarchos-implementer agent to implement this task using TDD in an isolated worktree." <commentary> Implementation task requiring test-first development triggers the implementer agent. </commentary> </example>
Use this agent when a task has failed and needs diagnosis and repair with adversarial verification. <example> Context: A delegated task failed its quality gates or tests user: "Task-005 failed TDD compliance — fix it" assistant: "I'll dispatch the exarchos-fixer agent to diagnose and repair the failure." <commentary> Failed task requiring root cause analysis and targeted fix triggers the fixer agent. </commentary> </example>
Use this agent when performing read-only code review for quality, design compliance, and test coverage. <example> Context: Feature implementation is complete and needs review user: "Review the agent spec handler for code quality" assistant: "I'll dispatch the exarchos-reviewer agent to analyze code quality and design compliance." <commentary> Code review request triggers the reviewer agent for read-only analysis. </commentary> </example>
Use this agent for low-complexity scaffolding tasks — file creation, boilerplate generation, and structural setup. <example> Context: Orchestrator needs new files or boilerplate created user: "Create the directory structure and stub files for the new feature" assistant: "I'll dispatch the exarchos-scaffolder agent to generate the scaffolding in an isolated worktree." <commentary> Simple file creation and boilerplate generation triggers the scaffolder agent with concise output. </commentary> </example>
Guided authoring of an architectural invariant catalog entry through a 6-step interview (elicit, locate, weight, enforce, number, commit). Drives the invariants_scaffold and invariants_add orchestrate verbs — the agent supplies judgment and natural-language elicitation; the verbs own schema validation, file writing, and event emission. Defaults to mode: audit; mode: check is an advanced opt-in. Triggers: 'add an invariant', 'author an invariant', 'enforce an architectural rule', or /exarchos:invariants. Do NOT use for: editing workflow state, running a review, or hand-writing YAML (the verbs write it — never emit catalog YAML yourself).
Collaborative design exploration for new features and architecture decisions. Triggers: 'brainstorm', 'ideate', 'explore options', or /ideate. Presents 2-3 approaches with trade-offs, documents chosen approach. Do NOT use for implementation planning or code review. Requires no existing design document — use /plan if one exists.
Post-merge workflow resolution. Verifies PR merge status, backfills synthesis metadata, force-resolves review statuses, transitions to completed, and cleans up worktrees/branches. Use when the user says 'cleanup', 'resolve workflow', 'mark as done', or runs /cleanup. Do NOT use before PRs are merged.
Bug investigation and fix workflow. Triggers: 'debug', 'fix bug', 'investigate issue', 'something is broken', or /debug. Hotfix track for quick fixes, thorough track for root cause analysis. Do NOT use for feature development or refactoring. Do NOT escalate to /ideate unless the fix requires architectural redesign.
Dispatch implementation tasks to agent teammates in git worktrees. Triggers: 'delegate', 'dispatch tasks', 'assign work', or /delegate. Spawns teammates, creates worktrees, monitors progress. Supports --fixes flag. Do NOT use for single-file changes or polish-track refactors.
Uses power tools
Uses Bash, Write, or Edit tools
Share bugs, ideas, or general feedback.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Portable Development System — AI-assisted development methodology with skills for consistency and agents for scale.
Long-running agent harness with 5-layer memory architecture, GitHub integration, autonomous batch processing, Agent Teams with ATDD, 9 hooks (safety, quality gates, team coordination), and 6 Agent Skills
Persona-driven AI development team: orchestrator, team agents, review agents, skills, slash commands, and advisory hooks for Claude Code
Corca Workflow Framework — consolidated hooks and skill orchestration for structured development sessions
Multi-agent orchestration framework for Claude Code, Gemini CLI, and Codex CLI — 19 agents, 13 skills, 15 commands, quality gates, TDD enforcement
Multi-agent /workflow development pipeline (planner → plan-review → coder → code-review) with typed handoff contracts, lifecycle hooks, and MCP servers.
Backend code quality skills for Claude Code. Nine skills that audit, critique, harden, simplify, humanize, and design-constrain your architecture — the backend half of impeccable.
Your agents forget. Exarchos doesn't.
Persistent SDLC state for any AI coding agent. Survives /clear, auto-compaction, and context overflow.
First-class with Claude Code, Codex, Cursor, OpenCode, Copilot; works with any agent that can run a CLI.
Install · What's different · What you get · Architecture · Docs
A plan.md per feature. CLAUDE.md rewritten between sessions. Summaries scrawled before /clear so the next session has something to start from. Phases enforced by you reminding the agent. It works. It's also manual, and one long context window away from the agent ignoring all of it.
/clearReturn to any suspended workflow by running /rehydrate.
❯ /exarchos:rehydrate payments-v2-migration
Workflow Rehydrated: payments-v2-migration
Phase: implementing | Type: feature
Task Progress
4 of 7 complete · last commit on feature/payments-v2
Artifacts
Design: docs/designs/payments-v2.md
Plan: docs/plans/payments-v2.md
PR: not yet created
Next Action
Continue task 5 (gates pending). Run /delegate or pick up manually.
State doesn't live in your conversation. It lives in an append-only event log. /rehydrate is a projection that rebuilds the workflow document for a fresh context window. The whole thing fits in about 2,500 tokens.
A state machine owns phase transitions, not a paragraph in CLAUDE.md. Convergence between phases ("is this implemented?", "does it match the design?") runs as TypeScript checks against your diff and git history, not prompts the agent can talk itself out of. You approve the design and you approve the merge. The middle runs on its own.
Run /ideate to start.
The CLI is the universal surface. Each runtime talks to it through whichever invocation it speaks natively.
| Runtime | Transport | Skill rendering | Slash commands |
|---|---|---|---|
| Claude Code | Plugin + MCP | First-class (rendered + hooks) | Yes (/ideate, /plan, etc.) |
| Codex CLI | MCP | First-class | Via Codex's command surface |
| Cursor | MCP | First-class | Via Cursor's MCP integration |
| OpenCode | CLI | First-class | Via OpenCode's runtime |
| GitHub Copilot CLI | CLI | First-class | Via Copilot's runtime |
| Anything else | CLI | Generic bundle | Whatever your agent supports |
The CLI works universally. For Claude Code, the recommended install path is the plugin.
Standalone CLI / MCP server (any agent, any runtime):
# Unix (macOS / Linux)
curl -fsSL https://lvlup-sw.github.io/exarchos/get-exarchos.sh | bash
# Windows (PowerShell)
irm https://lvlup-sw.github.io/exarchos/get-exarchos.ps1 | iex
exarchos --version
exarchos doctor
exarchos mcp // starts MCP server over stdio
exarchos install-skills
Auto-detects which runtime is on your PATH and installs the matching skill bundle. One match installs that bundle. Multiple matches prompt you to pick. No match installs the generic bundle and tells you what it found and why. Skip detection with --agent claude (or codex, opencode, copilot, cursor, generic).
/plugin marketplace add lvlup-sw/.github
/plugin install exarchos@lvlup-sw
Same binary underneath. Adds Claude Code slash commands, hooks, and rendered skills.
No SSH key? Use the HTTPS URL:
https://github.com/lvlup-sw/.github.git
For two-step (download + inspect + run), channel selection, validation, update, and uninstall: see the full install guide.
Other approaches in this space optimize for different things. None are wrong. They answer different questions.