{"id":47601652,"url":"https://github.com/marvinrichter/clarc","last_synced_at":"2026-04-01T18:51:26.448Z","repository":{"id":345551658,"uuid":"1174805635","full_name":"marvinrichter/clarc","owner":"marvinrichter","description":"Production-grade Claude Code workflow OS — agents, skills, hooks, commands, rules, and a continuous learning flywheel.","archived":false,"fork":false,"pushed_at":"2026-03-20T07:37:34.000Z","size":19539,"stargazers_count":5,"open_issues_count":1,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-03-20T07:43:39.609Z","etag":null,"topics":["ai-agents","claude","claude-code","cli","developer-tools","llm","productivity","workflow"],"latest_commit_sha":null,"homepage":"https://marvinrichter.github.io/clarc","language":"JavaScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/marvinrichter.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":".github/FUNDING.yml","license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":"SECURITY.md","support":null,"governance":null,"roadmap":"docs/roadmaps/ROADMAP.md","authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null},"funding":{"github":"marvinrichter"}},"created_at":"2026-03-06T21:31:16.000Z","updated_at":"2026-03-19T22:57:58.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/marvinrichter/clarc","commit_stats":null,"previous_names":["marvinrichter/clarc"],"tags_count":1,"template":false,"template_full_name":null,"purl":"pkg:github/marvinrichter/clarc","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/marvinrichter%2Fclarc","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/marvinrichter%2Fclarc/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/marvinrichter%2Fclarc/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/marvinrichter%2Fclarc/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/marvinrichter","download_url":"https://codeload.github.com/marvinrichter/clarc/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/marvinrichter%2Fclarc/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31290966,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-01T13:12:26.723Z","status":"ssl_error","status_checked_at":"2026-04-01T13:12:25.102Z","response_time":53,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai-agents","claude","claude-code","cli","developer-tools","llm","productivity","workflow"],"created_at":"2026-04-01T18:51:25.712Z","updated_at":"2026-04-01T18:51:26.441Z","avatar_url":"https://github.com/marvinrichter.png","language":"JavaScript","funding_links":["https://github.com/sponsors/marvinrichter"],"categories":[],"sub_categories":[],"readme":"# clarc\n\n**The workflow OS for Claude Code.**\n63 agents · 249 skills · 178 commands · 20 language rule sets · continuous learning flywheel.\n\n[![CI](https://github.com/marvinrichter/clarc/actions/workflows/ci.yml/badge.svg)](https://github.com/marvinrichter/clarc/actions/workflows/ci.yml)\n[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)\n[![Node.js 20+](https://img.shields.io/badge/node-%3E%3D20-brightgreen)](https://nodejs.org)\n\n---\n\nclarc turns Claude Code from a coding assistant into a structured engineering system. It brings together agents that delegate work, skills that encode domain knowledge, commands that run repeatable workflows, hooks that automate the background, and a learning loop that improves itself over time.\n\n---\n\n## The clarc Way\n\nclarc ships a complete, opinionated development methodology — from raw idea to shipped code.\n\n```\nPhase 0: Discovery        /idea → /evaluate → /explore → /prd\nPhase 1: Planning         /plan\nPhase 2: Implementation   /tdd  (RED → GREEN → IMPROVE)\nPhase 3: Quality          /code-review  +  /security\nPhase 4: Ship             /commit-push-pr\n```\n\n**Skip what you don't need:**\n\n| Task | Start here |\n|------|-----------|\n| New idea (unsure if worth building) | `/idea` |\n| Feature with clear spec | `/plan` |\n| Bug fix | `/tdd` (write the failing test first) |\n| Refactor | `/plan` → `/tdd` |\n| Docs / chore | `/commit` |\n\nRun `/clarc-way` to get a tailored recommendation for your current task.\nRun `/quickstart` if you're new to clarc.\n\n---\n\n## Getting Started in 5 Minutes\n\n```bash\n# 1. Install (pick your language)\nnpx github:marvinrichter/clarc typescript\nnpx github:marvinrichter/clarc python\nnpx github:marvinrichter/clarc go\n\n# 2. Verify the installation\nnpx github:marvinrichter/clarc doctor\n\n# 3. Open Claude Code and run your first command\n# /quickstart    → guided tour of the most important workflows\n# /clarc-way     → what to do for your current task\n```\n\n---\n\n## Install\n\n```bash\n# Recommended — interactive wizard (auto-detects your languages)\nnpx github:marvinrichter/clarc\n\n# Or with explicit language(s) — skips the wizard\nnpx github:marvinrichter/clarc typescript\nnpx github:marvinrichter/clarc typescript python go\n\n# Other editors\nnpx github:marvinrichter/clarc --target cursor typescript\nnpx github:marvinrichter/clarc --target opencode typescript\nnpx github:marvinrichter/clarc --target codex\n\n# Check if your installed rules are up to date\nnpx github:marvinrichter/clarc --check\n```\n\nThe wizard clones clarc to `~/.clarc/` on first run, then **symlinks** agents, commands, and rules into `~/.claude/`. A single `git pull` in `~/.clarc/` keeps everything current — no re-install needed.\n\n```\n~/.claude/agents/tdd-guide.md  →  ~/.clarc/agents/tdd-guide.md   (symlink)\n~/.claude/commands/plan.md     →  ~/.clarc/commands/plan.md       (symlink)\n~/.claude/rules/common/*.md    →  ~/.clarc/rules/common/*.md      (symlinks)\n```\n\nYour own agents or rules placed in `~/.claude/agents/` or `~/.claude/rules/` are never touched — symlinks only fill in files that don't already exist there.\n\n```bash\n# Update clarc\ncd ~/.clarc \u0026\u0026 git pull\n\n# Force copy instead of symlinks (CI, containers, cross-filesystem)\nnpx github:marvinrichter/clarc --copy typescript\n```\n\n\u003cdetails\u003e\n\u003csummary\u003eClone and run locally\u003c/summary\u003e\n\n```bash\ngit clone git@github.com:marvinrichter/clarc.git ~/.clarc\n~/.clarc/install.sh typescript\n~/.clarc/install.sh --enable-learning typescript\n~/.clarc/install.sh --check\n```\n\u003c/details\u003e\n\n---\n\n## What's inside\n\n| Component | Count | Purpose |\n|-----------|------:|---------|\n| **Agents** | 63 | Specialized subagents — delegate planning, review, testing, debugging |\n| **Skills** | 249 | Domain knowledge — patterns, conventions, examples for specific tasks |\n| **Commands** | 178 | Slash commands — repeatable workflows triggered by `/command` |\n| **Rules** | 20 | Language rule sets — always-on coding standards and checklists |\n| **Hooks** | — | Background automations — format, lint, persist state, weekly digests |\n\n---\n\n## Agents\n\nAgents are delegated automatically based on what you're doing. You can also invoke them explicitly.\n\n| Agent | When it activates |\n|-------|------------------|\n| `planner` | Complex features, refactoring, anything requiring a plan first |\n| `tdd-guide` | New features and bug fixes — enforces write-tests-first |\n| `code-reviewer` | Immediately after writing or modifying code |\n| `security-reviewer` | Before commits touching auth, input handling, secrets |\n| `architect` | Architectural decisions, ADR creation |\n| `build-error-resolver` | When the build fails |\n| `typescript-reviewer` | TypeScript-specific patterns and type safety |\n| `python-reviewer` | Python idioms, PEP 8, async patterns |\n| `go-reviewer` | Idiomatic Go, error handling, goroutine patterns |\n| `java-reviewer` | Spring Boot, JPA, hexagonal architecture |\n| `swift-reviewer` | SwiftUI, Combine, concurrency |\n| `ruby-reviewer` | Idiomatic Ruby, Rails best practices, RuboCop |\n| `elixir-reviewer` | OTP patterns, Ecto, Phoenix context violations |\n| `product-evaluator` | Go/No-Go evaluation of new feature ideas |\n| `solution-designer` | Generates 2–4 solution options with trade-off analysis |\n| `refactor-cleaner` | Identifies and removes dead code safely |\n| `doc-updater` | Keeps documentation in sync with code changes |\n| `android-reviewer` | Jetpack Compose, Hilt scopes, Room migrations, Coroutines |\n| `gitops-architect` | Designs GitOps setup — ArgoCD/Flux, env strategy, progressive delivery |\n| `finops-advisor` | Cloud cost analysis, rightsizing, ROI-prioritized recommendations |\n| `devsecops-reviewer` | SAST, secrets detection, IaC misconfigurations, OWASP Top 10 |\n| `mlops-architect` | ML serving stack, monitoring, drift detection, A/B testing plans |\n| `sdk-architect` | SDK architecture, multi-language generation, release process |\n| `frontend-architect` | Micro-frontend architecture, Module Federation, team topology |\n| `data-architect` | Data Mesh design, domain products, migration from monolith |\n| `docs-architect` | API documentation strategy, Mintlify/Docusaurus, Divio structure |\n\n---\n\n## Key commands\n\n### Daily workflow\n\n```\n/clarc-way     Interactive guide — recommends the right workflow for your task\n/quickstart    First-time onboarding — understand clarc in 5 minutes\n/plan          Plan implementation before writing code\n/tdd           Enforce test-first development\n/code-review   Security and quality review\n/verify        Full build, type check, lint, test cycle\n/build-fix     Incrementally fix build and type errors\n```\n\n### Product \u0026 architecture\n\n```\n/idea          Capture and structure a raw product idea\n/evaluate      Go/No-Go evaluation before building\n/prd           Write a Product Requirements Document\n/explore       Generate and compare solution options\n/breakdown     Break a feature into sprint-ready user stories\n/adr           Record an Architecture Decision\n/arc42         Generate or update arc42 architecture documentation\n```\n\n### Engineering operations\n\n```\n/slo                   Define SLIs, SLOs, and error budget alerts\n/instrument            Add analytics event tracking to code\n/experiment            Design a statistically valid A/B test\n/incident              Open and manage a production incident\n/deps                  Dependency audit and upgrade planner\n/setup-ci              Generate GitHub Actions CI/CD pipeline\n/security              Full DevSecOps scan — SAST, secrets, deps, DAST, OPA\n/gitops-review         Review GitOps config — sync, secrets, drift, progressive delivery\n/finops-audit          Cloud cost audit — tagging, rightsizing, anomalies, Infracost\n/zero-trust-review     Service mesh \u0026 mTLS review — identity, NetworkPolicies, east-west\n/iac-review            IaC code review — Pulumi/CDK/Terraform abstractions, compliance\n/mlops-review          MLOps audit — experiment tracking, serving, drift detection\n/privacy-audit         Privacy engineering audit — PII scan, retention, RTBF, consent\n/android-review        Android/Compose review — UDF, Hilt, Room, Coroutines\n/mobile-release        Mobile release workflow — signing, beta, store submission\n/webrtc-review         WebRTC architecture review — TURN, signaling, simulcast, security\n/frontend-arch-review  Micro-frontend architecture review — federation, routing, DX\n/data-mesh-review      Data Mesh review — domain boundaries, contracts, quality, lineage\n/sdk-review            SDK design review — ergonomics, backward compat, error design\n/docs-review           API documentation audit — completeness, playground, changelog\n/onboard               Generate onboarding materials — CONTRIBUTING.md, architecture tour\n```\n\n### Learning \u0026 evolution\n\n```\n/learn-eval         Extract reusable patterns with quality gate + save-location decision\n/evolve             Promote instincts into permanent skills\n/instinct-status    Show what the system has learned\n/sessions           Browse session history and snapshots\n```\n\n---\n\n## Skills\n\nSkills are loaded on-demand when Claude detects they're relevant. They encode domain knowledge in a structured format: when to use, patterns, examples, anti-patterns.\n\n**Language patterns:** TypeScript · Python · Go · Java · Rust · Swift · C++ · Ruby · Elixir · Django · Spring Boot · React Native · SwiftUI · Android/Jetpack Compose · Flutter · WebAssembly · Kotlin · Scala · PHP · R · C# · Bash\n\n**Architecture:** Hexagonal · DDD · Strategic DDD · API design · gRPC · GraphQL · WebSockets · Micro-Frontends/Module Federation · Zero-Trust/Service Mesh · Data Mesh · CQRS/Event Sourcing · Multi-Agent Systems\n\n**Data:** PostgreSQL · DuckDB · NoSQL · Redis/caching · database migrations · Data Mesh (Great Expectations, Soda, Delta Lake, Iceberg)\n\n**Infrastructure:** Terraform · Pulumi · AWS/Azure CDK · Kubernetes · Docker · GitOps (ArgoCD, Flux) · FinOps (Infracost, OpenCost) · CI/CD · deployment patterns · Edge/Serverless\n\n**Testing:** TDD workflow · E2E (Playwright) · contract testing (Pact) · visual regression · load testing · eval harness · chaos engineering · MLOps (A/B model testing, drift detection)\n\n**Security:** DevSecOps (SAST/DAST, OPA/Kyverno) · supply chain (SBOM/SLSA) · privacy engineering (PII, Presidio) · GDPR · auth patterns · Zero-Trust\n\n**AI/ML:** LLM app patterns · RAG · eval harness · MLOps · cost-aware pipelines · foundation models · prompt engineering · SDK design\n\n**Ops:** Observability · SLO workflow · incident response · performance profiling · platform engineering · developer onboarding\n\n**Product engineering:** Analytics workflow · experiment design · feature flags · A/B testing · compliance \u0026 audit logs · API documentation engineering\n\n**Team process:** OKRs · sprint planning · DORA metrics · roadmapping · working agreements\n\n---\n\n## Continuous learning flywheel\n\nclarc gets better the more you use it. There are two learning mechanisms:\n\n**Session learning** (`/learn-eval`)\nAt the end of a session, extracts recurring patterns, friction points, and solved problems into instinct files — with a quality gate and save-location decision (global vs. project-scoped) before writing. These become part of your personal context in future sessions.\n\n**Instinct evolution** (`/evolve`)\nAnalyzes accumulated instincts and promotes stable, high-signal ones into permanent skills. Closes the loop from observation → codified knowledge.\n\n**Weekly evolve batch**\nEvery Monday, clarc automatically analyzes accumulated instincts and generates a digest with promotion suggestions — without waiting for a manual `/evolve` call.\n\n**Conflict detection**\nWhen contradictory instincts exist in the same domain (e.g. \"prefer functional\" vs \"use classes\"), they are flagged in `conflicts.json` and surfaced in `/instinct-status` for resolution.\n\n**Pre-compact snapshots**\nBefore context compaction, clarc reads the session transcript and persists a structured snapshot: open tasks, touched files, recent exchanges. Nothing is lost between context resets.\n\n---\n\n## Language rules\n\nRules are guidelines that complement skills: rules define *what* to do, skills show *how*. `common/` rules are always active. Language rules activate automatically when you open files of that type (via `globs:` frontmatter) — they don't add overhead for projects that don't use them.\n\n**Performance tip:** install only the language sets you need to keep context lean.\n\n```\nrules/\n├── common/          # Universal — immutability, error handling, security, GitOps, onboarding\n├── typescript/      # TS/JS — strict typing, ESM, React patterns\n├── python/          # PEP 8, type hints, async idioms\n├── go/              # Idiomatic Go, error wrapping, goroutines\n├── swift/           # Swift 6, actors, value types\n├── java/            # Spring Boot, JPA, hexagonal architecture\n├── rust/            # Ownership, Result/Option, unsafe rules\n├── cpp/             # C++17/20, RAII, smart pointers\n├── ruby/            # RuboCop, Rails conventions, RSpec, Brakeman\n├── elixir/          # mix format, Credo, OTP patterns, Sobelow\n├── kotlin/          # Coroutines, null safety, sealed classes\n├── bash/            # set -euo pipefail, quoting, shellcheck\n├── scala/           # Functional idioms, ADTs, effect systems\n├── c/               # C11/C17, memory safety, opaque pointers\n├── php/             # PHP 8.4+, strict types, PSR-12\n├── r/               # tidyverse, native pipe, renv\n├── csharp/          # C# 12/.NET 8, nullable, records\n├── sql/             # Query optimization, indexing, migrations\n├── flutter/         # Widget lifecycle, state management, null safety\n└── android/         # Compose UDF, Hilt scopes, Room, Coroutines\n```\n\n---\n\n## Multi-editor support\n\nclarc ships configurations for multiple editors alongside the primary Claude Code integration:\n\n- **Claude Code** — agents, skills, commands, hooks, rules (full support)\n- **Cursor** — 96 language rules in `.cursor/rules/` + `.cursorrules` global config (`--target cursor`)\n- **OpenCode** — 65 commands + agent prompts under `.opencode/` (`--target opencode`)\n- **Codex CLI** — instructions + 14 commands under `codex/` (`--target codex`)\n\n### clarc as MCP server\n\nclarc can expose its own state as an MCP server, letting subagents and external tools query it directly:\n\n```json\n{ \"mcpServers\": { \"clarc\": { \"command\": \"node\", \"args\": [\"\u003cpath\u003e/mcp-server/index.js\"] } } }\n```\n\nAvailable tools: `get_instinct_status` · `get_session_context` · `get_project_context` · `skill_search` · `agent_describe` · `rule_check` · `get_component_graph` · `get_health_status`\n\n---\n\n## Structure\n\n```\nclarc/\n├── agents/          # Subagent definitions (frontmatter + instructions)\n├── skills/          # Domain knowledge skills (SKILL.md per topic)\n├── skills/INDEX.md  # Machine-readable skill catalog by domain\n├── commands/        # Slash commands (/tdd, /plan, /breakdown, ...)\n├── hooks/           # Hook configurations (JSON)\n├── scripts/hooks/   # Hook implementations (Node.js)\n├── rules/           # Language rule sets (20 languages)\n├── mcp-configs/     # MCP server configurations\n├── mcp-server/      # clarc as an MCP server (8 tools)\n├── .opencode/       # OpenCode commands and agent prompts\n├── .cursor/         # Cursor rules (50+ language rule files)\n├── codex/           # Codex CLI instructions and commands\n├── tests/           # Structural evals, behavior evals, unit tests\n└── install.sh       # Installer: --check, --target cursor/opencode/codex, --enable-learning\n```\n\n---\n\n## Based on\n\nclarc is a heavily extended fork of [everything-claude-code](https://github.com/affaan-m/everything-claude-code) by Affaan Mustafa, released under the MIT License.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmarvinrichter%2Fclarc","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmarvinrichter%2Fclarc","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmarvinrichter%2Fclarc/lists"}