Help us improve
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Run a structured multi-agent development pipeline (design → plan → review → code → review) enforced by typed handoff contracts, lifecycle hooks, TDD discipline, and code-quality gates — all within Claude Code.
npx claudepluginhub hex0xdeadbeef/claude-kit --plugin claude-kitImplements code strictly per approved plan
Explores requirements and designs approach before planning. Produces approved spec for /planner.
CRUD lifecycle manager for Claude Code artifacts — commands, skills, rules, agents. Use when user asks to "create command", "create skill", "enhance artifact", "audit artifacts", "delete rule", "rollback changes", or "onboard project". Modes: create, enhance, audit, delete, rollback, list, onboard. Features 9-phase workflow with quality gates, constitutional evaluation, and progress tracking. Do NOT use for writing application code, file operations, or general coding tasks.
Researches codebase and creates detailed implementation plan
Autonomous agent for deep project analysis and .claude/ configuration generation
Explores codebase to gather context for planning and implementation. Use when you need to understand existing patterns, find implementations, or analyze architecture before making changes.
Reviews code changes for architecture compliance, security, error handling, and test coverage. Use when code needs review before merge.
Reviews implementation plans for architecture compliance, completeness, and security. Use when a plan needs validation before coding begins.
Lightweight agent for recovering a verdict when code-reviewer or plan-reviewer fails to return one. Runs git diff, outputs VERDICT + brief handoff. No memory, no skills, no checklists.
Project-local terse-output mode for /workflow pipeline. Cuts Messages token cost via concise prose. Loaded only at SessionStart by .claude/scripts/caveman-activate.sh — not auto-invoked on user prompts.
Review standards for code-reviewer agent. Loaded by explicit Read in the code-reviewer STARTUP step (disable-model-invocation blocks subagent preload). Covers: severity classification (BLOCKER/MAJOR/MINOR/NIT), decision matrix (APPROVED/APPROVED_WITH_COMMENTS/CHANGES_REQUESTED), auto-escalation rules, grep search patterns for automated checks.
Implementation rules and patterns for /coder command. Load at /coder startup (step 0) or when /workflow enters Phase 3. Covers: 5 CRITICAL rules (plan-only, import matrix, clean domain, no log+return, tests pass), evaluate protocol (PROCEED/REVISE/RETURN), dependency-ordered implementation.
Design phase rules for /designer command. Load at /designer startup (step 0). Covers spec quality, design checklist, approach evaluation criteria.
Review standards for plan-reviewer agent. Loaded by explicit Read in the plan-reviewer STARTUP step (disable-model-invocation blocks subagent preload). Covers: architecture compliance (import matrix, domain purity), required plan sections validation, severity classification (BLOCKER/MAJOR/MINOR/NIT), decision matrix (APPROVED/NEEDS_CHANGES/REJECTED).
Matches all tools
Hooks run on every tool call, not just specific ones
Executes bash commands
Hook triggers when Bash tool is used
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.
This plugin requires configuration values that are prompted when the plugin is enabled. Sensitive values are stored in your system keychain.
provision_settings_localOn first session in this project, merge the kit's opinionated env defaults (prompt-cache TTL, git co-author strip, auto-memory, strict-mode toggles) into .claude/settings.local.json — your existing values always win. Equivalent to what install.sh does for a project-scoped install. Off by default; effective the next session.
${user_config.provision_settings_local}Production-ready Claude Code configuration with role-based workflows (PM→Lead→Designer→Dev→QA), safety hooks, 44 commands, 19 skills, 8 agents, 43 rules, 30 hook scripts across 19 events, auto-learning pipeline, hook profiles, and multi-language coding standards
PROJECT.md-first autonomous development with hybrid auto-fix documentation. 8-agent pipeline, auto-orchestration, docs auto-update on commit (true vibe coding). Knowledge base system with 90% faster repeat research. Strict mode enforces SDLC best practices automatically. Works for ANY Python/JavaScript/TypeScript/Go project.
Implementation planning, execution, and PR creation workflows with multi-agent collaboration
Describe your goal, approve the spec, then step away — Claude and Codex loop together until it's right.
Corca Workflow Framework — consolidated hooks and skill orchestration for structured development sessions
Verification-first engineering toolkit for Claude Code. 15 skills across a 5-phase spine (Investigate → Design → Implement → Verify → Ship), 8 specialist agents, an interactive setup wizard. Every skill has rationalizations + evidence requirements. Built for senior ICs and tech leads.
Modifies files
Hook triggers on file write and edit operations
Modifies files
Hook triggers on file write and edit operations
Uses power tools
Uses Bash, Write, or Edit tools
Uses power tools
Uses Bash, Write, or Edit tools
Share bugs, ideas, or general feedback.
Claude Kit
Reusable configuration kit for Claude Code
English | Русский
Structured multi-agent development workflow with built-in planning, implementation, and code review phases. Supports any language and framework — Go, Python, TypeScript, Rust, Java, and 26 more via tree-sitter analysis.
Note: Defaults are tuned for Go (sparse paths,
pre-commit-build.shrunsgo build,CLAUDE.mdLanguage Profile pins Go ≥ 1.24). For other stacks: editCLAUDE.mdLanguage Profile, adjustworktree.sparsePathsin.claude/settings.local.json, and replace or disable the Go-specific build hook. See ⚙️ Configuration Files.
Requirements: Claude Code
>= 2.1.141(the kit's hooks rely on the exec-formargs:invariant from v2.1.139 andterminalSequenceJSON output from v2.1.141). On earlier versions hooks silently no-op / degrade.
/workflow orchestrates planning → implementation → review → commit for any task. Onboarding in four steps.
Install claude-kit as a native Claude Code plugin: shared across all your projects, versioned, updated through the marketplace, with nothing copied into your repo. This repo doubles as its own marketplace (.claude-plugin/marketplace.json + .claude-plugin/plugin.json).
/plugin marketplace add hex0xdeadbeef/claude-kit
/plugin install claude-kit@claude-kit
Plugin commands are namespaced — you run /claude-kit:workflow, /claude-kit:planner, /claude-kit:coder, /claude-kit:designer. The internal pipeline delegation (planner → plan-reviewer → coder → code-reviewer) resolves automatically by description, so it works regardless of the prefix.
install.sh (project-scoped — copies the kit into one repo)A project-scoped install copies .claude/ + CLAUDE.md into the repo, so you can customize rules / Language Profile per project and commit them with your team. One command provisions everything — the .claude/ pipeline, the 3 MCP servers (.mcp.json), and personal settings (.claude/settings.local.json):
curl -sL https://raw.githubusercontent.com/hex0xdeadbeef/claude-kit/main/install.sh | bash
No manual cp needed — settings.local.json and .mcp.json are created automatically (on --update, new defaults merge in while your edits are preserved). The MCP servers need npx (Node.js) and/or uvx (uv); if either is missing the installer prints the install command. After installing a runtime, restart Claude Code — the servers auto-load on next start (verify with claude mcp list).
Plugin vs install.sh — both coexist:
CLAUDE.md, architecture rules, .claude/PROJECT-KNOWLEDGE.md).install.sh — project-scoped: the kit lives in your repo, customizable and committed with your team./project-researcher # writes .claude/PROJECT-KNOWLEDGE.md
/workflow Add new REST endpoint for profiles
/project-researcher gives the agents context for your codebase (architecture, modules, dependencies, language profile). Edit the CLAUDE.md Language Profile first if your stack is not Go (the kit's default). Then /workflow drives the full cycle: task analysis → [design — L/XL only] → planning → plan review → implementation → code review → commit. (In plugin mode, prefix the commands: /claude-kit:project-researcher, /claude-kit:workflow.)
The kit is tuned with environment variables (validation strictness, prompt-cache TTL, terse-output mode, and more). Where you set a variable decides its scope — and the mechanism is identical for the plugin and the install.sh install: Claude Code injects the env block into the session, and every hook script (plugin hooks included) inherits it as a subprocess.