AWESOMECLAUDECODE

A code-intelligence Claude Code plugin and MCP server (Rust, single binary, 43 languages) that builds a typed dependency graph and packs token-budgeted, annotated context bundles for any task. Exceptionally engineered and security-forward — 2,400+ tests, recorded ADRs, a sandboxed WASM plugin SDK, and cosign-signed, SBOM-attested release images.

A mature MCP server (plus a 33-endpoint REST API) that drives Google NotebookLM for citation-backed Q&A and full Studio generation (audio, video, infographics, reports), with multi-account TOTP re-auth. Notably well-maintained and security-attentive — batch-tested on overnight 1,000-question runs, with a documented changelog that includes patching a transitive XSS advisory.

A Claude Code plugin that turns CSS selectors + text into finished visual documentation — annotated screenshots, flow GIFs/MP4s, terminal recordings, and before/after composites — placing every annotation deterministically and pixel-verifying each artifact (PASS/FAIL) before it is saved. Self-contained (bundles its own headless Chromium, no browser MCP, no telemetry); every image in its README is generated by the tool itself.

A multi-agent Claude Code learning environment for mastering Retrieval-Augmented Generation, with 20 specialist agents, 22 slash commands, and a 9-module hands-on curriculum that runs zero-config inside Claude Code. Quality is evident — 616 tests, CI, and weekly/monthly content-freshness automation that opens issues for stale material.

A thorough, device-tested guide and toolkit for running Claude Code natively on Android via three paths (Termux, proot-Ubuntu, and the Android Virtualization Framework), with a verification harness and per-device results. Unusually security-aware for a setup guide — it ships a threat model, an SSRF-guard WebFetch hook, a biometric approval gate, and a permission matrix.

A self-contained Claude Code skill that turns Benjamin Bloom's 2-sigma tutoring research into a personal AI tutor: it generates a structured syllabus, teaches one lesson at a time, and adapts each next lesson to the learner's annotations and feedback. The skill is dependency-free and makes no network calls beyond the configured LLM endpoint (an optional web app is also included).

A Claude Code skill plus single-file HTML tool that closes the human-review loop on AI-written docs: a person leaves inline comments in the browser, which export as structured JSON keyed by heading path and line, and the skill polls for that file and applies each comment to the exact lines. Strongly privacy-respecting — the local/CLI build enforces a strict CSP (`connect-src 'none'`), so document text and comments never leave the machine.

A collection of installable thinking-framework skills with a meta-router, notable for publishing a replication-gated evaluation instead of unsupported quality claims.

A research-lab systematic analysis of the Claude Code codebase whose headline finding — overwhelmingly infrastructure rather than model — reframes Claude Code as a harness.

A mirror of the Anthropic© PBC documentation site for Claude/Code, but with bonus features like full-text search and query-time updates - up-to-the-minute, fully-indexed information so that Claude Code can read about itself.

A comprehensive and well-documented collection of agentic patterns from Anthropic docs, with colorful Mermaid diagrams and code examples for each pattern. Covers Subagent Orchestration, Progressive Skills, Parallel Tool Calling, Master-Clone Architecture, Wizard Workflows, and more. Also compatible with other providers.

Collection of best practices, tips, and techniques for Claude Code development workflows, enhanced with distributable plugins

This is either the work of a prolific genius, or a very clever bot (or both), although it hardly matters because the quality is so good - an index of 75+ Claude Code repositories published by the author - and I'm not talking about slop. CMS, system design, deep research, IoT, agentic workflows, server management, personal health... If you spot the lie, let me know, otherwise please check these out.

All parts of Claude Code's system prompt, including builtin tool descriptions, sub agent prompts (Plan/Explore/Task), utility prompts (CLAUDE.md, compact, Bash cmd, security review, agent creation, etc.). Updated for each Claude Code version.

A freely available reference covering 190+ patterns for AI-assisted software development (and actually a whole bunch of related technical topics) from foundational concepts through agentic construction patterns, governance, testing, and socio-technical systems. Each entry follows a consistent pattern-language format with Context, Problem, Forces, Solution, Consequences, and Related Patterns. Opinionated and erudiate, which is actually good for an encyclopedia in some ways.

A creative educational framework for Claude Code, inspired by the FASTER approach to self-teaching. Ships with a variety of agents, slash commands, and tools that enable Claude Code to help you progress at your own pace, employing well-established pedagogical techniques like active learning and spaced repetition.

A Claude Code skill that turns any codebase into a beautiful, interactive single-page HTML course for non-technical vibe coders.

A multimodal, grounded active-learning tutor engine for college textbooks. Mitigates AI hallucinations and diagram blindness with automated vector figure cropping, homework-solution auto-pairing, and cross-session persistent state. Benchmarked on STEM courseware to run on small models with significant token savings.

Skills for knowledge work rather than just coding, covering a variety of cognitive approaches to problem-solving: brainstorming, critical thinking, problem diagnosis, design thinking, market research, deep reading, Socratic tutoring, and more.

A catalog of 15 Claude Code skills mapped to 8 research-workflow stages (literature → gap analysis → design → drafting → reviewer response), where each stage emits an explicit YAML/Markdown deliverable the next stage consumes. Thoughtfully designed around anti-hallucination — schemas force gap status on unsupported claims, and downstream skills refuse malformed, overconfident handoffs.

A ready-to-fork Claude Code template for academics using LaTeX/Beamer + R. Multi-agent review, quality gates, adversarial QA, and replication protocols. Great use of orchestration patterns and great documentation.

A set of ready-to-use Agent Skills for research, science, engineering, analysis, finance and writing. That's their description - modest, simple. That's how you can tell this is really one of the best skills repos on GitHub. If you've ever thought about getting a PhD... just read all of these documents instead. Also I think it IS an AI agent or something? Awesome.

An agent skill that geolocates a photo with no readable text. Single-purpose Python scripts search and rank candidates using OpenStreetMap geometry, elevation skylines, satellite imagery and street view. The answer comes with a camera position, heading, error radius and an evidence image. NOTE: Scripts involve contact with Baidu and Yandex.

Skills for geospatial analysis that require coordinate-system, spatial-validation and comparability checks before a result is reported, and narrow or refuse a claim the data cannot support. Ships a paired evaluation against a skills-disabled control arm.

A local-first router that sits under Claude Code (and Codex/Gemini CLI) and sends each prompt to the cheapest capable model, with three-layer token compression and automatic provider fallback — protecting your premium quota and cutting cost with zero config on a Claude subscription. Credibly engineered: 1,900+ tests, an independent RouterArena benchmark placement, and an honest local-only design with no hosted proxy.

An agent-native skill that accesses 90+ websites by calling their underlying APIs directly (typed JSON in, JSON out) instead of screenshotting and parsing the DOM, with auth auto-resolved locally from your existing browser session. Security is first-class — every operation is tagged read/write/delete/transact behind permission tiers, SSRF protection runs on each request, and there is no vendor backend or proxy in the request path.

A zero-dependency Claude Code skill (68 commands) that connects to your *real* Chrome — logged-in tabs, cookies, live page state — to give the agent a perception layer: layout, visible styles, per-action what changed evidence, CSS-cascade-to-source tracing, and session replay/Playwright export. Notably rigorous about its own claims, with a dogfood benchmark gate that blocks performance/adoption assertions unless it passes.

Converts a running Express or FastAPI app into an MCP server by AST-parsing its routes and injecting a marked, byte-for-byte-removable `/mcp` router, so the agent can operate your live application (real auth, real data) rather than just read files. Exceptionally safety-minded: writes are disabled by default behind two-phase confirmation, docstrings are sanitized against prompt injection, a circuit-breaker quarantines failing routes, nothing leaves the machine, and persistence stores structure never payloads.

A tiny bridge that lets Claude Code drive desktop software — Photoshop, Premiere, Blender, Unity, InDesign, Office, 13 apps total — by writing one-time scripts against each app's own automation runtime (COM / AppleScript / CEP), instead of MCP servers or fragile screenshot-based computer use. A genuinely novel approach that unlocks the apps' full scripting surfaces (InDesign alone exposes ~28k objects) with low maintenance.

A focused collection of documented scripts that fix the most painful Claude Code papercuts on WSL2 + Windows Terminal — clipboard screenshot paste via a Go daemon, Windows notifications on Stop/PermissionRequest hooks, LSP wiring, a usage-aware status line, and voice-mode audio routing. Each fix is written up with root cause and exact config (including non-obvious traps like SessionEnd firing per subagent), and the work was upstreamed into Awesome-WSL.

Enables users to prompt codex from claude code. Unlike the raw codex mcp server, this skill infers parameters such as model, reasoning effort, sandboxing from your prompt or asks you to specify them. It also simplifies continuing prior codex sessions so that codex can continue with the prior context.

A well-organized, well-written set of plugins covering core developer activities, such as working with common cloud platforms like GitHub, Azure, MongoDB, and popular services such as Tavily, Playwright, and more. Clear, not overly-opinionated, and compatible with a few other providers.

Fuses 2–5 frontier models instead of racing them — one architect, one primary builder, and up to three secondary builders — giving N-way opinions, debate, fusion, coordinated implementation, direct one-agent routing, and gate-first validation. Ships as a composable Pi extension with pre-built model stacks and install/prime commands, and a video walkthrough of the whole approach.

Local, multi-provider agent communication tool. Adds hooks and a skill to exchange questions and reviews, and saves decisions, verification results and handoffs for later sessions. Each session keeps its own client, model, conversation and permissions.

A live mesh network that connects Claude Code and other providers so they can talk to each other in real time. Built for emergent coordination. Agents discover each other and share context across project boundaries. Uses tmux sessions as the coordination layer with logical circles for peer isolation. Includes a web dashboard and Telegram bot.

Runs Claude Code fully on-device on Apple Silicon: a native MLX server speaks the Anthropic Messages API, so the unmodified claude CLI works against local Qwen, Gemma and Llama models with no cloud calls. Includes RAM-based model selection, a tool-call parser for local models' output formats, and a verified-offline mode.

Streams a locally-running Claude Code session live into a Slack or Mattermost thread so a whole team can watch, type, and approve actions together — screen-sharing for AI pair programming, but everyone can type. Mature and well-maintained (npm-published with CI and coverage), with reaction-based tool approvals, git-worktree isolation, multi-account rotation, and per-thread sessions that survive restarts.

A zero-dependency, cross-platform notifier that fires a desktop toast and a free phone push (via ntfy) the moment Claude Code (or Codex/Cursor/Gemini) finishes a task or needs input, wired up by a one-command setup. The standout among notifiers is its testing rigor — CI drives the *real* agent CLIs end to end and asserts real ntfy round-trips and real OS toast delivery, no mocks.

A feature-rich Telegram bot that turns your machine into a remote Claude Code terminal driven from your phone: streaming tool progress, voice input, a git panel, Mac remote control, and 30+ commands. Its standout is the approval flow — dangerous-op Approve/Deny taps are injected straight into the real terminal prompt via `tmux send-keys`, so you authorize from your phone without killing the session or losing context.

Local, offline voice-to-text for Claude Code on Linux built on faster-whisper, with a warm daemon for instant transcription, system-wide push-to-talk, a voice-enabled editor, and per-project vocabulary hints that fix technical terms (Claude not cloud). Fully private — no cloud, no API keys, with thoughtful touches like mic-health monitoring and the ability to offload transcription to a LAN GPU; hardware-tested down to a Jetson Orin.

A native-Swift macOS menu-bar app (10 MB, no Electron) that covers and input-locks your screen with one hotkey while your agents keep running, then makes the locked screen glow when Claude Code pauses for permission or finishes. Polished and well-engineered — CI with 50 tests, signed/notarized builds, an honest visual privacy tool, not a security boundary disclosure, no analytics, and an idempotent hook installer that backs up and never clobbers foreign hooks.

Connects WhatsApp as a native Claude Code channel via Baileys linked-device (no bot token or API keys), with bidirectional messaging, full media, voice transcription, remote tool approval, access control, and per-group personalities. Runs entirely locally and was the first community plugin officially reviewed and published on Anthropic's plugin marketplace.

A no-frills little library that adds delightful OS-native sounds to Claude Code via simple hooks. It really sparks joy.

A web and mobile PWA for driving Claude Code (and Cursor/Codex/Gemini) from any device — file explorer, git, integrated shell, and full session management that reads and writes your real `~/.claude` config rather than duplicating it. By far the most-adopted Claude Code UI, with a self-hostable open-source core, an optional Docker microVM sandbox mode, a plugin ecosystem, and tools disabled by default for safety.

A cross-platform desktop IDE built on an infinite zoomable canvas, where editors, terminals, browsers, and docs float in freeform space instead of tabs — and ships skills that let Claude Code spawn agent terminals on the canvas, coordinate through canvas notes, and drive browser panels. A genuinely novel UX, well-engineered (CI, e2e tests, context-isolated IPC, scoped filesystem access) and widely adopted.

A cross-platform desktop IDE that wraps Claude Code sessions with a swarm mode (parallel agents in a grid), a real-time session inspector (cost, tokens, tool-usage, context), multiple isolated Claude profiles, a kanban board, and encrypted P2P session sharing over WebRTC. Signed/notarized, broadly adopted, and built squarely for the multi-agent, multi-session workflow.

A visual workspace for building with Claude Code (and Codex) where you and the agent co-edit *visually* — markdown, mockups, mermaid, Excalidraw, CSV, and data models — approving the agent's changes as red/green WYSIWYG diffs, with session/task kanban, worktrees, an extension SDK, and a native iOS companion app. A distinctive higher-bandwidth take on agent collaboration, open source with transparent, opt-out telemetry.

A native Linux GUI for Claude Code agents built in Rust + GTK4 — single binary, no Electron, starts in under a second — with a file tree, syntax-highlighted viewer, markdown preview, embedded VTE4 terminal, streaming agent chat, workspace tabs, and tool-restricted agent profiles. Fills a real gap as essentially the only native (non-webview) Linux desktop client, using your own Claude Code CLI with no proxy.

A VS Code extension and standalone terminal dashboard that adds visibility and AI conveniences on top of your Claude Max subscription — inline completions, code transforms, AI commit messages, plus deep session observability (token-burn, a 13-week quota heatmap, multi-account management, cross-session search, and asset extraction). Mature and well-maintained (CI, published to the marketplace and npm, extensive docs).

A frameless, always-on-top floating chat window for Claude Code on Windows. It captures every monitor and lets Claude read the screen to answer questions in context, and drives the user's existing claude CLI login through the Agent SDK, with image paste, model switching, live streaming, and a context-usage statusline.

Spawn and control multiple Claude Codes in parallel from your phone or desktop. Happy Coder runs Claude Code on your hardware, sends push notifications when Claude needs more input or permission, and costs nothing. End-to-end encrypted, with realtime voice.

Local web UI that runs the unmodified Claude Code CLI in a pseudo-terminal and adds tabs, a browsable history of past sessions, conversations rendered as cards, a context meter, and the agent's multiple-choice questions answered from the UI.

The most complete Claude Code status line: 5-hour and 7-day rate-limit usage with reset countdowns and *learned* end-of-window projections, context window, prompt-cache-expiry countdown, per-session cost, plus live todo/tool/git activity — across 3 styles and 9 themes, configurable by CLI or natural-language skill. Genuinely well-engineered (PyPI, 320+ tests, a sub-1%-CPU daemon, deep docs) and the clear standout of a very crowded field.

A delightfully different status line: 30+ kaomoji text-faces that react in real time to what Claude is doing — context-aware personas by file type, and a frustration-escalation system where mounting errors progress from `( ദ്ദി ˙ᗜ˙ )` toward a table-flip `(╯°□°)╯︵ ┻━┻`. Pure Rust with sub-2ms rendering, a single binary, hook-based activity tracking, and an interactive config TUI — the standout creative pick in a sea of metrics bars.

Shows each Claude Code session's state in the tmux tab bar, per window: an animated spinner while it works, amber when it waits on input, red on a permission request, green when done. Event-driven via hooks with no polling; catches unreported transitions (Esc, permission grants, stop-hook continuations) with transcript watchers. 260+ test assertions run in CI.

A Go CLI that gives each Claude Code session its own creature in the status line, derived by hashing the session id, with one den per git worktree. Where other status-line faces react to what the agent is doing, this one is a stable identity for telling parallel sessions apart, and its expression tracks repository hygiene: uncommitted files, unpushed commits, distance behind trunk and the last test result.

A really stacked status line that exposes just about everything you might need - context usage, tools, agents, todos, etc. Highly configurable and actively maintained at the time of writing - code quality is strong.

A highly customizable status line formatter for Claude Code CLI that displays model info, git branch, token usage, and other metrics in your terminal.

A vim-style powerline statusline for Claude Code with real-time usage tracking, git integration, custom themes, and more.

A high-performance Claude Code statusline tool written in Rust with Git integration, usage tracking, interactive TUI configuration, and Claude Code enhancement utilities.

Tracks the carbon footprint of Claude Code sessions, feautring a live CO2 estimate in the status line, computed locally from token usage, next to the session cost.

A browser-automation plugin/skill that lets Claude Code drive a browser to test and verify its own work — full Playwright API plus pixel- and DOM-level computer-use toolsets, connecting to your running Chrome or a fresh Chromium. Notably secure and fast: scripts execute inside a QuickJS WASM sandbox with no host filesystem or network access, and a published benchmark shows it beating Playwright MCP on time, cost, and turns.

A design engine that takes a different tack from feed the model more tokens: it teaches design *judgment* — ~74 rules pros carry but never write down (the refined black isn't #000, it's #2A2A2A; one accent color, everything else grayscale) — as plain markdown the agent reads automatically, plus a brand-agnostic skin system, a named motion vocabulary, and `/ss-*` review skills. Widely adopted and demonstrably effective (a live demo morphs one UI across Toss/Raycast/Arc via a single attribute).

A deep design-engineering skill that makes agents design like they have taste by default, layered so you can just install it, drive it with 22 single-lens commands, or wire its deterministic MCP gates and CLI into CI. Its signature is a *scoreable, defensible* critique — Nielsen's heuristics × classic design laws × persona walkthroughs, every finding tagged by business impact — backed by anti-slop rules, outcome recipes, accessibility-first review, and 31 domain references.

A visual whiteboard between you and your agent: Claude renders diagrams, HTML, or UI components through Snip instead of describing them in text, you approve or annotate directly on the output (circle, arrow, note), and the agent gets structured feedback and iterates. Works via CLI or MCP, doubles as a full local screenshot/annotation app with on-device AI organization (Ollama), and adds a genuine human-in-the-loop visual review step to Claude Code.

Agent skill and plugin that turns complex terminal output into styled HTML pages or slide decks for diagrams, diff reviews, plan audits, data tables, and project recaps. Generates self-contained browser-readable artifacts with Mermaid diagrams, responsive layouts, themes, and sharing commands.

Another welcome contribution to the domain of making Claude Code output look stylish, this collection of skills produces a variety of editorial diagrams in the form of self-contained HTML + SVG. Although I never thought I'd hear these words together, according to the author: no Mermaid-slop.

Easily generate web assets from Claude Code including favicons, app icons (PWA), and social media meta images (Open Graph) for Facebook, Twitter, WhatsApp, and LinkedIn. Handles image resizing, text-to-image generation, emojis, and provides proper HTML meta tags.

Create and evolve SVG logos with Claude Code. Turns a project brief into distinct concepts you can compare side by side, then guides refinement, checks favicon legibility, and exports production-ready PNGs.

A CLI and Claude Code skill for UI debugging in Chrome using the DevTools Protocol, including DOM inspection, browser events, rendered styles, and detailed layout information.

Claude Code plugin with eight design-automation skills driven by one schema-validated brand.json: design tokens for CSS, Tailwind, TypeScript and Figma, social cards, favicon and app-icon sets, responsive AVIF and WebP images, HTML-to-video recording, a WCAG 2.1 AA audit, and a runtime UX audit covering bfcache, INP, dialogs and mobile viewports.

Local MCP server, Claude Code hooks and a Chrome extension for marking up a running app: clicking an element (or one in the iOS Simulator) and typing a change delivers the comment to Claude Code with the selector, DOM path, computed styles, component chain, source-file hint and a cropped screenshot.

Agent [Skill](https://github.com/i18next/i18next-cli/tree/main/skills/i18next-localization) that takes an app from hardcoded strings to a localized one with i18next: stack detection, config, AST-based wrapping of strings in `t()`, key extraction, sync to Locize and AI translation. It regenerates the step-by-step prompt from the installed i18next-cli, so it cannot drift from the CLI.

A portable writing skill that audits and rewrites prose to remove AI-isms — 49+ pattern categories and a tiered word-replacement vocabulary, with detect / rewrite / edit-in-place modes, content-type and voice profiles, and a two-pass re-check. Stands well above the crowded de-slop field: it ships a deterministic, zero-dependency detector engine (the single source of its 0–100 score) with CI tests, and frames itself honestly as a writing-quality tool, not a detector-evasion gimmick.

A Claude Code skill for naming products, SaaS tools, brands, and projects via a structured metaphor-driven process — naming brief, metaphor exploration, candidate generation, anti-slop filtering, a weighted evaluation rubric, and availability checks for domains/handles/package names. A genuinely different (and creative) take that produces names grounded in meaning instead of the random tech-syllable mush an unguided model reaches for.

A comprehensive pipeline of Skills that replicates traditional publishing infrastructure for nonfiction book creation using specialized Claude skills.

A CLAUDE.md section that constrains Claude's prose style. It bans specific habits, including announcing a point before making it, colon-hinged sentences where the left side labels the right, verbless fragment openers, and stacked compression. Each rule names the habit, shows an example, and gives the rewrite.

Six Agent Skills for scientific writing based on Stanford's Writing in the Sciences course: sentence-level prose editing, IMRaD manuscript structure, the writing process, peer review and publication, grants and recommendation letters, and lay summaries and press. Ships as the sciwrite Claude Code plugin with four stdlib-only Python audit scripts.

A plugin for turning technical ideas into patent disclosures.

Skills for planning and drafting fiction as markdown with YAML frontmatter: story bible, characters, worldbuilding, arcs, scenes, and chapters. Includes a dependency-free CLI whose continuity checker flags characters appearing after their death, payoffs placed before their setup, and stale story state.

A zero-dependency Node CLI (and Claude Code plugin/skill) that lets the agent edit CapCut / JianYing video projects programmatically — inspect timelines, build drafts, add text/audio, word-level captions and Whisper transcription, templates, and cut long-form into shorts — JSON in, JSON out, no server. Deeply mature for a media tool: 205 tests across macOS/Windows/Linux CI, a huge documented command surface, and a deliberately stateless the CLI is the protocol design.

An open-source collection of ~50 motion-graphics, animation, and video skills across 14 installable packs — kinetic typography, data-driven charts, explainers, TikTok/Reels, web/WebGL animation, and Manim math animation — that teach an agent how a professional would build each piece. Every visual skill ships a deliver-and-verify loop (render a frame → screenshot → check) plus a small verify toolkit, so the agent inspects its own output instead of guessing at timing and framing.

Vox Director is an open-source skill that turns a one-line topic into a finished Vox-style paper-collage explainer or ad video. It automates the full pipeline — script, collage keyframes, motion, voice-over, music, and captions — on the Atlas Cloud API plus local ffmpeg, with two human approval gates for the beat map and visual style.

An outstanding, creative library that converts Claude Code session transcripts into self-contained, embeddable HTML replays - interactive playback with speed control, a local editor, collapsible tool-call and thinking blocks, redaction for private information - the output is fantastic. Also supports other providers.

A Claude Code plugin that drives Blender over an MCP server to turn a text brief or a reference photo into an optimized, validated GLB: scripted modeling measured against the photo, textures from CC0 scans, a joint-by-joint review, Draco/WebP export.

A best-practices skill that teaches the agent to write safer Terraform and OpenTofu through a diagnose-first workflow, failure-mode routing, LLM-mistake checklists, and a feature-version guard table mapping features to their version floor and common errors — covering testing, modules, remote state, CI/CD, and security scanning across AWS/Azure/GCP. From an AWS Hero behind the terraform-aws-modules ecosystem, and the most authoritative, most-adopted IaC resource in the set.

A focused, vendor-neutral skill that writes, fixes, and *validates* OpenTelemetry Collector configs — encoding the Collector's real footguns (memory_limiter/batch ordering, core-vs-contrib components, pull-vs-push exporters, deprecated exporters, pipelines that validate but were never wired) as an authoring workflow and diagnosis checklist. Its standout is honesty plus a verification loop: it always finishes by running `otelcol validate` and ships CI-checked before/after example configs rather than handing back an unverified file.

Immensely detailed set of skills for DevOps Engineers (or anyone who has to deploy code, really). Works with validations, generators, shell scripts and CLI tools to create high quality IaC code for about any platform you've ever struggled painfully to work with. Worth downloading even just as a source of documentation.

Packages a PWA into a signed Play Store Android bundle with Bubblewrap and checks target API level, signing key reuse, assetlinks fingerprints and version codes before upload.

Agent skills and a Claude Code plugin marketplace for mirrord, which runs a local process inside a live Kubernetes cluster's network, environment and traffic. Lets Claude verify changes against real services, databases and queues instead of mocks, with skills for config, CI, database branching, Kafka queue splitting and preview environments.

Security scanner for AI agent skills. Detect vulnerabilities, malicious patterns, and security risks.

Sandboxed devcontainer for running Claude Code, Codex, and OpenCode in bypass / auto-approve mode.

The Execution Security Layer for the Agentic Era. Providing deterministic Sudo governance and audit logs for autonomous AI agents.

A coding agent CLI hook that acts as a safety net, catching destructive git and filesystem commands before they execute. Supports Codex, Claude Code, OpenCode, Gemini CLI, Copilot CLI, Kimi Code and Pi.

One isolated Lima VM per GitHub project — sandboxed Claude Code/Codex, Docker, Node, signed git

Pin AI Skill behavior. Block unapproved drift in CI. See exactly what changed in every PR.

CLI tool for running coding agents inside hardware-isolated microVMs

3-level override system for Claude Code - prevents destructive system operations. Born from a real incident.

Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically..

Developer-grade Claude Code + Codex configuration: cost-tiered subagents, workflow commands, guardrail hooks, MCP parity, and an installable plugin/marketplace.

Give the agent a cage, not your keys. One-command Docker sandbox for AI coding agents: full autonomous permissions, per-project isolation, your host stays untouched.

Real-time secret-leak guardrails for AI coding agents (Claude Code, Codex), Git hooks, and CI.

Credential gateway for AI agents. Log in once via Oauth2 or API Key. Every agent stays authenticated — headless, no SaaS, agents never see your credentials.

Airut is a system for running Claude Code tasks from email and Slack. It handles workspace provisioning, container isolation, network sandboxing, session persistence, and cleanup — a secure foundation for autonomous agentic development.

Runtime guardrails for Claude Code. Auto-approve what's safe, gate what's risky, block what's dangerous. Dual enforcement, full audit trail. MIT.

Containerized development environments for coding agents, from the Dagger team. Gives each agent its own isolated stack so several can work in parallel without reaching into your host or each other's workspaces.

A very professional collection of over a dozen security-focused skills for code auditing and vulnerability detection. Includes skills for static analysis with CodeQL and Semgrep, variant analysis across codebases, fix verification, and differential code review.

Auto-approve safe bash commands using AST-based parsing, while prompting for destructive operations. Solves permission fatigue without disabling safety. Supports Claude Code, Gemini CLI, and Cursor.

Prompt injection scanner for AI coding tool hooks, written in Rust. Catches injection attacks, leaked secrets, and data exfiltration in tool inputs and outputs by running DeBERTa v3 and optionally Meta Llama Prompt Guard 2 locally via Candle. Built for Claude Code hooks; the models are gated on HuggingFace, so installation requires accepting their licences first. Self-described as early development.

Gryph is a security layer for AI coding agents. It hooks into Claude Code and enforces YAML policy rules that block, warn, or guide actions in real time, then logs every event to a local SQLite database for review and audit.

Claude Code marketplace for multi-agent collaboration — task splitter, output reconciler, adversarial debate, shared memory, acceptance gate. Composes with codex-delegate / gemini-delegate.

Garry Tan's (Y Combinator) Claude Code setup and open source software factory for managing the development lifecycle end-to-end. Includes a set of agents and in-depth skills/tools along with workflows for advancing a product from ideation to production.

A Claude Code development harness that enables reliable high-quality development through an autonomous Plan -> Work -> Review cycle. Well documented and includes an Output Style.

A principled, spec-driven workflow that transforms large problems into focused, incremental missions using Claude Code's specialized sub agents. Includes slash-commands, sub agents, and specialized workflows designed for specific parts of the SDLC.

A meta-skill that designs domain-specific agent teams, defines specialized agents, and generates the skills they use. Resources are in Korean but can produce high-quality English-language output.

A set of commands that provide a comprehensive workflow system for managing projects, including task management, code review, and deployment processes.

Structured development workflow enforcing separation between Research, Innovate, Plan, Execute, and Review phases. Features consolidated subagents for context-efficiency, branch-aware memory bank, and strict mode enforcement for guided development.

Relays messages between two Claude Code sessions on separate machines using the Channels API, with delivery tracking and per-message reply budget. Pairs sessions belonging to different accounts, which native cross-session messaging does not. Although this sort of functionality is now a platform feature, it's always cool to see people who have built their own solutions.

A versatile configuration framework that enhances Claude Code with specialized commands, cognitive personas, and development methodologies, such as Introspection and Orchestration.

A terminal app that manages multiple Claude Code, Codex, and other local agents (including Aider) in separate workspaces, allowing you to work on multiple tasks simultaneously.

A very pragmatic set of well-designed agents, skills, and commands, built around a discipline of turning past mistakes and errors into lessons and opportunities for future growth and improvement. Good documentation.

Workflow automation system for Claude with a group of useful plugins, agents, and skills. Automates task-to-production workflows, PR management, code cleanup, performance investigation, drift detection, and multi-agent code review. Includes [agnix](https://github.com/avifenesh/agnix) for linting agent configurations. Built on thousands of lines of code with thousands of tests. Uses deterministic detection (regex, AST) with LLM judgment for efficiency. Used on many production systems.

Really comprehensive and feature-packed project-management workflow for Claude Code. Numerous specialized agents, slash-commands, and strong documentation.

Three review workflows developed at an AI-native startup: automated PR code review (dual-loop, GitHub Actions plus slash command), security review against OWASP Top 10 with severity-classified findings, and the widely-copied design-review agent that drives Playwright MCP to check UI/UX and accessibility against a live preview. Each ships as a slash command, a subagent, and a workflow file, with video walkthroughs.

Interactive review UI that intercepts ExitPlanMode via hooks, letting you visually annotate plans with comments, deletions, and replacements before approving or denying with detailed feedback. Has since grown to cover code and diff review, PRs, and rendered HTML artifacts, with feedback sent straight back to the agent. Supports several coding agents alongside Claude Code.

Lightweight CLI tool for real-time communication between Claude Code sub agents using hooks. Enables multi-agent collaboration with @-mention targeting, live dashboard monitoring, and a zero-dependency implementation.

Combines Claude Code, Codex, and Pi terminal sessions into persistent long-running teams. Agents message each other directly over tmux and queue work to each other, instead of the user relaying messages. Teams and agents are saved as YAML files, and a running team can be snapshotted and restored later.

A self-hosted office with a pleasing pixel-art UI where one manager agent, Billion, takes a mission, posts jobs to a board, and reviews and merges pull requests that Claude Code and Codex workers open from their own git worktrees. Accessible from desktop or mobile.

Spec-first, test-driven agent framework: runs the build loop against live Lakebase Postgres branches under a deterministic orchestrator, with human-approval gates.

One local, searchable index of your AI coding-agent history Claude Code, Codex, Cursor, Gemini & more. Keyword + semantic search, MCP server, CLI & VS Code extension.

Always-on companion for Claude that remembers your decisions and their evolution. Local-first memory using SQLite + transformers.js embeddings.

High-fidelity transcript memory for Claude Code — index every session, recall byte-identical, search, prune, compose. Local-first, stdlib-only, MCP.

Per-repo memory, outcome telemetry, and a calibrated-confidence gate for Claude Code, with MCP and AGENTS.md projections so other AI coding tools can read its context. Notes survive sessions; success claims need test evidence; your reverts are remembered. Local-only, stdlib runtime.

Outcome-based persistent memory MCP server for Claude Code and OpenCode. Good advice promoted, bad advice demoted. pip install roampal.

Hivemind turns your traces into reusable skills across agents

Memory management and retrieval for Claude Code

🦫 Privacy-first virtualization layer for LLM context with MCP protocol for tool access.

Long-term memory for AI-coded codebases. A git blame for AI agents — but for the why. MCP server that captures the agent's reasoning live, in context, as each change is made. Local SQLite, zero deps.

Claude Code plugin that tracks token usage, identifies wasted context, and saves money on unnecessary API costs, by tracking which information is actually being reused later. Visuals include heatmaps, ROI reports, budget alerts, efficiency scores, git-aware suggestions - all local, zero config. The design is still somewhat exploratory, but it shows promise.

Authors and versions CLAUDE.md and AGENTS.md from a repository's actual detected stack, so Claude Code's project context stays in sync with the code instead of drifting. One typed source file (project.faf, an IANA-registered media type) is the single origin for every assistant's context file; a deterministic score reports how complete that context is. The file lives in the repo and carries no telemetry.

Hand-crafted collection of advanced context engineering techniques and patterns with minimal token footprint focused on improving agent result quality.

Open, local-first engram memory for Claude Code. Engrams are plain-text YAML you can read, correct, and delete. The same memory is shared across MCP tools and multiple providers.

Durable cross-agent memory for Claude Code, Codex, pi, and Hermes. Your memory is a Hugging Face dataset you own: recall it from any machine or share it with your team. Single binary, local index, pinned local embedding models.

Desktop workbench that reads Claude Code session files locally and renders the subagent and workflow topology as a live graph, with per-node latency, token cost, and nesting depth. Token spend is broken down by source (base session, skills, subagents, MCP) and can be traced from a cost bucket to the exact message that produced it.

Reads the JSONL session logs Claude Code already writes and shows, live during a turn, what the session is doing: each model call and tool, the context window filling, every subagent on its own window and its own model, and the commits and files the session produced. Bun + TypeScript server serving a local browser GUI, with an optional Tauri menu-bar tray; single-file binaries per platform, read-only and local-only.

Reads Claude Code JSONL session transcripts and draws a project's history as a creatively-styled diagram: a day, the tasks inside it, and every prompt, with the ones that ended in a commit marked. Segments work from pauses in the record rather than anything the user labelled, and prompts can be searched by the words they contain.

A pixel-art office for Claude Code agents, including a desktop widget and browser app, with tokens, a shop and a lounge. Friday is Hawaiian shirt day.

The linter and LSP for AI coding assistants — validates CLAUDE.md, AGENTS.md, SKILL.md, hooks, and MCP config, with autofixes and IDE plugins.

A language-agnostic linter (Rust) that keeps co-dependent code, docs, and config in sync, with a Claude Code plugin skill.

A CLI linter for AI agent context files that catches stale references, dead commands, and hardcoded secrets, with a modular tested rule set.

Deterministic quality scorer for AI agent instruction files — 8-dimension scoring with security, multi-format (SKILL.md, CLAUDE.md, .cursorrules, AGENTS.md), anti-gaming detection, zero dependencies

The tested, tool-agnostic AGENTS.md kit — verified templates, a CI linter, and migrators from .cursorrules/CLAUDE.md/Copilot/Windsurf/Cline/Aider.

Upkeep — an AI audit crew for your repo. Catches docs/spec/asset drift with evidence; output-only. Claude Code plugin/skill + reusable CI workflow.

Anthropic's official repository for Agent Skills — the SKILL.md format, a skill template, and example skills, the same format Claude Code loads natively.

Anthropic's official, curated directory of high-quality Claude Code plugins, installable from within Claude Code.

The official GitHub Action for running Claude Code in CI: mention @claude in issues and pull requests to delegate code changes, reviews, and fixes.

An official AI-powered security-review GitHub Action that uses Claude to analyze pull-request diffs for vulnerabilities.

Anthropic's canonical guide to working effectively with Claude Code: the agentic-loop mental model, CLAUDE.md guidance, and workflow patterns.

The official conceptual explainer of Claude Code's agentic loop, tools, and context window, and how skills, hooks, and subagents layer on top.

A framework for choosing which extension mechanism to reach for, organized around deterministic-versus-probabilistic control and context isolation.

Anthropic's foundational taxonomy of agent patterns — prompt chaining, routing, orchestrator-workers, and evaluator-optimizer — and when to use each.

A practical account of orchestrator and subagent coordination, prompt design, and evaluation that maps directly to Claude Code's subagents and agent teams.

Anthropic's guide to curating the context window — compaction, just-in-time retrieval, and note-taking — the discipline underlying effective long-horizon agent use.

Anthropic's official Claude Code cheatsheet — a quick reference for the core vocabulary (session, context window, CLAUDE.md), built-in slash commands, and keyboard shortcuts.

A current single-page reference for Claude Code: install, environment variables, slash commands, MCP, hooks, and subagents, kept in sync with the official changelog.

A thorough walkthrough of every hook event, when each fires, the two return channels, common anti-patterns, and copy-ready settings.json examples.

An essay on CLAUDE.md craft: instruction-budget reasoning, progressive disclosure, and the test of whether Claude would err without a given line.

A structured, chapter-based getting-started guide for Claude Code with a self-assessment quiz and a ten-module progressive learning path — slash commands, memory, skills, subagents, MCP, hooks, plugins, and checkpoints — with visual diagrams and copy-paste templates.

A conceptual, mental-models-first primer that explains what Claude Code is and how its agentic loop works, then layers setup, prompt-engineering workflows, skills, hooks, MCP, subagents, and agent teams, with an experience-tiered path for newcomers.

An interactive click-through of an annotated Claude Code project where every file and folder — CLAUDE.md, settings.json, rules, commands, skills, agents, hooks, plugins, and .mcp.json — is a real, explained concept, teaching the tool's surface area by orientation rather than prose.

A drop-in CLAUDE.md distilling four behavioral guidelines for LLM-assisted coding into Claude Code — a low-friction quick win. Karpathy-inspired, derived from Andrej Karpathy's public notes on LLM coding pitfalls and authored by multica-ai.

A really interesting analysis of how coding agents like Claude Code are designed. It attempts to break an agent down into its fundamental parts and reconstruct it with minimal code. Great learning resource. Final product is a rudimentary agent with skills, sub-agents, and a todo-list in roughly a few hundred lines of Python.

Really solid, insightful guidance on working/thinking with Claude Fable, and with AI in general. Very well written. Hints of Rumsfeld epistemology, but otherwise it's a great piece.

This has what you need. Remarkably clear, information-dense, it's Claude Code: the good parts, for beginners, advanced users, pets, anybody.

A Claude skill that activates Fable-style agentic behavior: explicit multi-stage planning, sub-agent delegation, and self-verification.

A plugin that conserves message tokens by communicating in fragmented caveman speak - sort of a clever form of compression. Now accompanied by a whole caveman ecosystem including a memory system, caveman spec kit, and a caveman agent.

A remarkably innovative approach to working with Skills, the centerpiece of which being a technique that leverages hooks to ensure that Claude intelligently selects and activates the appropriate Skill given the current context. Well-documented and adaptable to different projects and workflows.

A strong bundle of core competencies for software engineering, with good coverage of a large portion of the SDLC - from planning, reviewing, testing, debugging... Well written, well organized, and adaptable. The author refers to them as superpowers, but many of them are just consolidating engineering best practices - which sometimes does feel like a superpower when working with Claude Code.

A nice package of Claude Skills focused on SEO workflows: page audits, content briefs, article writing, E-E-A-T audits, semantic gap analysis, and more. Each skill fetches the target page and reads the top-ranking competitors itself, so users don't paste in keyword exports or crawl reports.

Useful skills and plugins based on daily.dev platform which contains daily curated news from trusted sources. Your agent can pull from curated list of news and real developer insights from 1M+ community.

Skills and an MCP server that let AI agents work with [Cooklang](https://cooklang.org/) recipes: plain-text .cook recipes and .menu meal plans in a folder you own.

Reviews Claude's uncommitted changes in a local GitHub-style pull request page: comment on any line, send comments in batches while Claude keeps working, ask Claude questions answered inline in the thread, and get each comment back marked fixed or refused.

A collection of board-specific Claude Code skills for microcontrollers firmware development, one per board, covering STM32, ESP32, ESP8266, RP2040/RP2350, nRF52840 and AVR.Each skill documents pin mapping, clock and peripheral pitfalls, PlatformIO/toolchain setup and flashing.

A set of skills for LinkedIn marketing: drafting posts, comments and replies, extracting the hook formula from a viral post, planning a week of content, rewriting a profile, and segmenting the people who engaged. Posts are read through Apify and published through Publora, and every skill shows a draft for approval before anything goes out.

End-to-end open source contribution manager. Discovers contributable issues across GitHub, tracks PRs across multiple repos, diagnoses CI failures, and drafts maintainer responses. Great collection of skills and agents for GitHub contributors, plus a sleek dashboard, standalone CLI and MCP server.

Skills for assessing and enhancing software projects to meet enterprise-grade standards for security, quality, and automation. Strong collection covering a range of development frameworks, enterprise-readiness, security, GitHub release management, branding, code review - really useful and well curated.

Command-line tool to customize your Claude Code installation: themes, thinking verbs, spinners, and input-box styling, plus deeper tweaks like custom toolsets, system-prompt edits, input pattern highlighters, and AGENTS.md support. Works against both native and npm installs.

A Node.js CLI tool that automatically generates configs (rules, ignore files, MCP servers, commands, and subagents) for various AI coding agents. Rulesync can convert configs between Claude Code and other AI agents in both directions.

A single appendable system prompt that retunes Opus 5's communication channel — cutting verbal tics, heading theater, and output-token bloat — passed via `--append-system-prompt-file` with no build step or dependencies. Comes with a `just`-driven side-by-side compare loop so you can see the difference, and a set of diagrams breaking down the prompt's anatomy. Opinionated, unusually specific about what it is fixing and why, and a good worked example of system-prompt engineering as a discipline.

Zsh completion for the Claude Code CLI. Dynamically completes sessions, MCP servers, agents, models and plugins from the user's configuration.

A CLI and TUI for running several Claude Code accounts on one machine. Each account gets its own config directory through a shell hook, keeping its own sign-in and history, while plugins, settings, skills, commands and agents are symlinked to one primary directory.

Well-crafted toolset for session continuity, featuring skills/commands to avoid compaction and recover context across sessions with cross-agent handoff between Claude Code and Codex CLI. Includes a fast Rust/Tantivy-powered full-text session search (TUI for humans, skill/CLI for agents), tmux-cli skill + command for interacting with scripts and CLI agents, and safety hooks to block dangerous commands.

Top-notch, well-written resources covering just about everything from core engineering domains. What's nice about this everything- store is most of the resources have significant standalone value and unlike some all-encompassing frameworks, although you can opt in to the author's own specific workflow patterns if you choose, the individual resources offer exemplary patterns in (just about) every Claude Code feature you can find (apologies to the Output Styles devotees).

A comprehensive Claude Code plugin with 65 specialized skills covering full-stack development across a wide range of specific frameworks. Features 9 project workflow commands for Jira/Confluence integration and, notably, an interesting approach to context engineering via a `/common-ground` command that surfaces Claude's hidden assumptions about your project. This is a smart thing to do.

A well-balanced, down-to-Earth set of sub agents, skills, and commands, that are well-organized, easy to read, and a healthy focus on meta-skills/agents, like skill-auditor, hook creation, etc. - the kind of things you can adapt to your workflow, and not the other way around.

Professional development environment for Claude Code with spec-driven workflow, TDD enforcement, cross-session memory, semantic search, quality hooks, and modular rules integration. A bit heavyweight but feature-packed and has wide coverage.

Comprehensive E2E development workflow with helpful Claude Code subagent prompts for solo devs. Run multiple auditors in parallel, automate fix cycles with micro-checkpoint protocols, and do browser-based QA. Includes strict protocols to prevent AI going rogue.

A hooks-driven system that monitors file operations in real-time and blocks changes that violate TDD principles.

A community list of Claude Code mods built with function hooks. Scans public GitHub repositories and shows what the validator reports each mod can read, write, run or send over the network.

Colorful, themeable Claude Code replies: tables, code, diagrams, charts and tool rows in over a dozen themes. Enhances mod development experience.