Tag
#claude-code
25 open-source projects filed under this tag.
hivemind
Hivemind is a shared memory layer for AI coding agents that captures traces, codifies reusable skills, and propagates knowledge across your entire team in real time.
addyosmani-agent-skills
Production-grade engineering skills for AI coding agents by Addy Osmani — 23 structured workflows for Claude Code, Cursor, Gemini CLI, and more.
adhd-agent
ADHD is a tree-of-thought reasoning skill for AI coding agents — spawns parallel isolated reasoning branches, prunes traps, and deepens the best ideas.
architect-loop
Claude Fable 5 as architect, GPT-5.5 Codex as builder — a research-backed cross-vendor agent loop skill for Claude Code with parallel worktrees and frozen gates.
beads
Beads is a Go-powered distributed graph issue tracker that gives AI coding agents persistent structured memory — replacing messy markdown plans with dependency-aware task graphs.
builderio-skills
Composable agent skills from BuilderIO for Claude Code and Codex — orchestration, visual planning, budget management, and code review workflows.
caveman
Caveman is a Claude Code skill that cuts 65% of AI output tokens by making agents talk like caveman — same accuracy, fewer words, lower cost. 70K GitHub stars.
claude-devtools
Visual debugging tool for Claude Code — inspect session logs, tool calls, token usage, subagents, and context window in one UI. 3.5K+ stars, MIT license.
context-mode
Context-mode is an MCP server that optimizes context windows for AI coding agents — 98% reduction in tool output size, session continuity, and support for 15+ platforms.
deepclaude
Swap Claude Code's brain for DeepSeek V4 Pro or OpenRouter — same autonomous agent loop, 17x cheaper. MIT licensed, 2K+ GitHub stars.
ecc
ECC is a harness-native operator system for agentic work — 209K stars, 251 skills, 63 agents across Claude Code, Codex, Cursor, Gemini, and more.
claude-bughunter
Claude-BugHunter is a 71-skill bundle for Claude Code that turns your AI agent into a senior bug-hunting researcher with 681 disclosed-report patterns across 24 vulnerability classes.
flock
DuckFlock is an open-source Go service that turns Telegram into an AI dev team — 5-role pipeline, per-chat workspaces, PR review, and Docker deployment.
gsd-core
GSD Core is a context-engineering framework for AI coding agents that prevents context rot through fresh subagents and spec-driven development workflows.
gstack
Garry Tan's open-source Claude Code setup turns AI into a 23-person virtual engineering team — CEO, designer, QA, security, and release engineer as slash commands.
harness
Harness is a Claude Code plugin that auto-generates agent teams and skills from domain descriptions — six architecture patterns, one prompt.
last30days-skill
AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and 10+ platforms — then synthesizes a grounded brief scored by real engagement.
loop-engineering
Loop Engineering provides practical patterns and CLI tools for designing systems that orchestrate AI coding agents — stop prompting, start engineering loops.
make-pages-interactive
A Claude Code skill that turns static HTML pages into live commenting surfaces — highlight text, leave notes, and Claude edits the page automatically.
munder-difflin
Munder Difflin is a local multi-agent harness for Claude Code that turns terminal sessions into a self-coordinating team with a visual office floor and GOD orchestrator.
ponytail
Ponytail is an AI agent skill that enforces YAGNI principles — cutting 80-94% of generated code by making your agent think like a lazy senior dev.
re_gent
re_gent is a Go-based version control system for AI coding agents that tracks every edit, prompt, and conversation — giving developers git-level auditability over agent activity.
regent
Regent is version control for AI coding agents — track every edit, blame prompts instead of humans, and audit what Claude Code or Codex did to your codebase.
sandbox-runtime
Anthropic's open-source sandbox runtime for AI agents — enforce filesystem and network restrictions at the OS level without containers. Built for Claude Code, works everywhere.
vibecode-pro-max-kit
Spec-driven coding harness for AI agents that kills context rot — 12 specialized agents, 32 skills, and self-improving memory for Claude Code, Codex, and Cursor.