Help us improve
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Complete Claude agentic dev-kit: 11 specialized subagents, 9 skills, 4 commands, 9 hook scripts, and 6 MCP integrations for the full software development lifecycle.
npx claudepluginhub atalias-mkyle/claude-dev-kit --plugin dev-kitSet up Claude Code in the current project — drops a small CLAUDE.md and .claude/memory/decisions.md tailored to the stack
Run the pre-PR self-review checklist before pushing
Re-run project-bootstrap in refresh mode — re-detects the stack and updates only the sections of CLAUDE.md that have drifted
Run a security review on the current branch or a given PR
Use when a CI run on a PR or branch has failed and you need to understand why before deciding whether to fix, re-run, or escalate.
Use when there is a bug to diagnose, a test is failing and the cause is unclear, or a runtime error needs reproducing before fixing.
Use to execute an approved plan. Receives the plan from the planner (or directly from the user) and makes the code changes in an isolated context, returning a diff summary at the end. The implementer does not improvise scope — if the plan is wrong, it stops and asks instead of making it up.
Use at session end when significant architectural decisions were made, or when .claude/memory/decisions.md is growing large or may contain stale/contradictory entries.
Use when starting a non-trivial task from scratch, when you need the full dev pipeline run end-to-end, or when an agent returns blocked/partial and you need recovery routing.
Use when something broke after a deploy, a production incident is active, or a rollback needs to be evaluated quickly.
Use when onboarding a new team member, starting work after a long break, or when asked "how do I get started" or "what's the state of this project".
Use before opening a pull request, when the user says "ready to ship", "let's PR this", "check before I push", or asks for a self-review pass. Walks through a pragmatic pre-PR checklist that catches the things reviewers always catch, so reviewers can focus on the things only humans catch.
Use when the user wants to set up Claude Code on a new project, asks to "create a CLAUDE.md", "bootstrap claude", "initialize claude for this repo", or when there is no CLAUDE.md in the current project root yet and the user is starting work. This skill drops in a small, project-specific CLAUDE.md based on what's actually in the repo, and creates a `.claude/memory/decisions.md` for architectural decisions.
Use when the user asks for a security review, security check, or wants to audit changes before merging — especially changes that touch auth, input handling, external API calls, file I/O, shell commands, or infrastructure configuration.
Executes bash commands
Hook triggers when Bash tool is used
Modifies files
Hook triggers on file write and edit operations
External network access
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.
memory_pathPath for the local knowledge graph memory store. Defaults to ~/.claude/dev-kit-memory.
${user_config.memory_path}brave_api_keyOptional. Enables web search for the researcher agent. Get a free key at brave.com/search/api.
${user_config.brave_api_key}memory_systemWhich memory system Claude should use to store and recall context. Options: openviking (recommended — auto-recall on every prompt), local-graph (mcp-knowledge-graph, lightweight JSONL store), decisions-md (no MCP, append to .claude/memory/decisions.md only).
${user_config.memory_system}context7_api_keyGet a free key at https://context7.com/dashboard for higher rate limits. Leave blank to use the free tier.
${user_config.context7_api_key}enable_autoformatRun prettier/black/gofmt/rustfmt after Claude writes or edits a file (if installed). Override per-session with DEVKIT_DISABLE_AUTOFORMAT=1.
${user_config.enable_autoformat}enable_bash_guardBlock rm -rf, force-pushes to protected branches, chmod 777, and similar. Also gates the write guard. Override per-session with DEVKIT_DISABLE_BASH_GUARD=1.
${user_config.enable_bash_guard}openviking_api_keyAPI key for your OpenViking server.
${user_config.openviking_api_key}openviking_base_urlBase URL of your OpenViking server. Defaults to http://127.0.0.1:1933 (standard local port). Override for a remote server.
${user_config.openviking_base_url}enable_session_contextOn SessionStart, inject git branch, recent commits, uncommitted changes, open TODOs, and recent decisions. Override per-session with DEVKIT_DISABLE_SESSION_CONTEXT=1.
${user_config.enable_session_context}Harness-native ECC plugin for engineering teams - 64 agents, 262 skills, 84 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
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.
Reliable automation, in-depth debugging, and performance analysis in Chrome using Chrome DevTools and Puppeteer
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.
Connects to servers outside your machine
Connects to servers outside your machine
Requires secrets
Needs API keys or credentials to function
Requires secrets
Needs API keys or credentials to function
Uses power tools
Uses Bash, Write, or Edit tools
Uses power tools
Uses Bash, Write, or Edit tools
Share bugs, ideas, or general feedback.
An opinionated, cross-machine Claude Code setup that you install once and get everywhere.
It bundles:
orchestrator, researcher, planner, reverse-reasoner, implementer, reviewer, security-reviewer, test-runner, debugger, ci-triager, memory-curator — each in its own context windowproject-bootstrap, debug-playbook, pr-checklist, security-review, dependency-audit, incident-response, deploy-check, onboard/bootstrap, /pr-check, /security-check, /refreshDesigned so installing it on a new machine is two commands.
You need these on each machine. Anything missing means a piece of the kit silently no-ops, never breaks.
| Tool | Why | Install |
|---|---|---|
python3 (3.10+ required) | Hook scripts. 3.9 will fail silently on session_context.py. Check: python3 --version | macOS: brew install python · Ubuntu: apt install python3 · Windows: python.org |
node + npx | Context7 + Sequential Thinking MCPs | nodejs.org or nvm |
uv (uvx) | Serena MCP | curl -LsSf https://astral.sh/uv/install.sh | sh (Mac/Linux) · winget install astral-sh.uv (Windows) |
git | SessionStart hook reads git state | already on most systems |
Optional but recommended formatters (auto-format hook detects what's installed): prettier, ruff or black, gofmt, rustfmt, shfmt, rubocop.
From inside Claude Code (any project, anywhere):
/plugin marketplace add atalias-tech/claude-dev-kit
/plugin install dev-kit@dev-kit-marketplace
You'll be prompted for the userConfig values (Context7 API key — optional, leave blank for free tier; OpenViking URL and API key; the boolean toggles).
That's it. On the next session start in any project, the hooks fire, the MCPs come up, and the subagents are available.
In any project, run:
/bootstrap
It drops a CLAUDE.md and .claude/memory/decisions.md tailored to whatever stack it detects. Review the file, fix the bits it got wrong, commit. Run /refresh later when the stack changes.
The kit ships three memory layers, active by default:
OpenViking is ByteDance's open-source context database for AI agents. The kit wires it directly — no separate plugin needed.
What the kit does automatically:
<openviking-context> so Claude starts every session with persistent context.<relevant-memories> before Claude responds.openviking MCP server registered at http://127.0.0.1:1933 by default, exposing mcp__openviking__search, mcp__openviking__store, mcp__openviking__read, mcp__openviking__forget, and others.Setup: Start an OpenViking server locally (docker run -p 1933:1933 volcengine/openviking) or point openviking_base_url at a remote one. Set openviking_api_key if your server requires auth. The kit writes ~/.openviking/ovcli.conf on first session start.
For cross-machine sharing, point OpenViking's data dir at a synced folder (~/Dropbox/openviking).
The session_capture.py Stop hook appends a timestamped git snapshot to .claude/memory/decisions.md at the end of every session. Syncs naturally if the project is on a synced drive.
If you prefer a local JSONL graph over OpenViking, set memory_path in userConfig and use the mcp__memory__* tools. Set memory_system: local-graph to have the session-start nudge point Claude at the right tools.