Recently added

Review playbook for Claude Code hooks: event/matcher selection, path resolution, script skeleton, fail-open/fail-closed posture, dispatch drift, the review checklist, and operational-footprint escalation to staff-platform-engineer. Install at project scope in repos that author .claude/hooks/*.sh scripts.

초보 바이브코더를 위한 AI 진입점 — 계획 먼저·쉬운 검토·안전 기본 (Claude Code + Codex)

Operator console for an Aeon agent instance: enable/schedule/edit skills, wire secrets and channels, debug runs, and mine past Codex/Claude Code chats into scheduled skills.

Монтаж записей Zoom-эфиров в готовые YouTube-видео: транскрибация Whisper локально, план монтажа, раскладки, слайды, плашки, субтитры и главы.

Read-only, deterministic PostgreSQL diagnostics as an agent tool: health, slow queries, unused indexes, bloat, vacuum, config — findings-first, never writes.

A bundle for product docs and the tasks that come out of them. Write a spec against a format and verify it before it ships; turn a request thread into a task's context table; file that task in your engineering tracker and keep the two sides reconciled. Where things are stored is decided by config.

A bundle for the half of Figma work that isn't drawing the screen itself. It builds your to-draw list before you start, and keeps a file several people share in order afterwards.

Read and explore .help/ documentation with progressive depth. Lightweight runtime companion to attune-ai's built-in authoring — no AI keys required.

Persistent memory and receipt-verified workflows for Claude Code — /spec plans, reviews, and executes with quality gates; plus security audits, code reviews, test generation, and release prep from /attune.

Contract gold fixture for plugin-gold profile validation.

Contract gold fixture for flat claude-plugin profile

27 bundled skills, smart worktree hooks, statusline, and MCP server stubs — plan, implement, review, debug, multi-agent setup, and audit code from one plugin.

Portable OrchestKit workflows for Codex: brainstorm, explore, implement, assess, verify, review pull requests, and render inline diagrams with glyph.

The Complete AI Development Toolkit — 109 skills, 36 agents, 171 hooks for full-stack development.

Portable OrchestKit workflows for Codex: brainstorm, explore, implement, assess, verify, review pull requests, and render inline diagrams with glyph.

The Complete AI Development Toolkit — 109 skills, 36 agents, 171 hooks for full-stack development.

Ready-to-use scientific and research Agent Skills for biology, chemistry, medicine, and related workflows.

Automated upgrade of legacy content, projects, and repositories: assess whether a legacy target is worth upgrading; scan for outdated dependencies, deprecated APIs, legacy directory structures, and non-standard fragments; plan, execute, and verify upgrades; roll back failed batches to baseline; batch-queue many legacy repositories from a CSV/JSON manifest (serial or bounded concurrency, failure retry, interrupt resume) and produce a unified summary report.

workset training companion — workout planning, set-by-set logging, and progress context for opengym2.

Read-only GitHub portfolio maintenance — repository readiness, branch risk, delivery hygiene, security posture, and ranked action plans via MCP.

Kroger/QFC shopping — product search, cart, shopping lists, pantry, weekly deals, and meal-planning context via MCP.

以時序制衡約束 agent——意圖揭露＋六段流程（requirement→research→plan→test→build→verify，intent 是意圖揭露的機械載體、流程之因，六段是實現意圖的手段鏈），兩層兩段式：定義層（理論）與實作層（實際）各成循環、段間切換一律 sb next 旗標切段；老闆任何新輸入（含兩時點 fail）回意圖揭露經 intent 路由器路由（定義級同 ms 開新輪）；審核資源集中在兩時點——時點 1＝requirement→research 邊（審意圖→需求翻譯）、時點 2＝build→verify 邊（審驗收資格——GWT 回指、假綠燈），對抗在前、老闆判定在後，pass 才 --boss-ok 推進；verify＝真驗收執行（GWT 逐條＝驗收劇本實操、行為證據落回指區），驗收完成老闆看行為證據終審 pass 出口（--new-ms 開下一里程碑或 sb end 結束 slug），時點對抗採 sb adversarial --point 1|2＋adversarialLog point 條目對照（雙流模型機械抗上下文壓縮），中鏈與段內提交零審核（提交僅 sb commitmsg 格式閘——單功能單提交）；RAM/ROM 分層（G/SLUG＝ROM 收斂產出、tmp/flow-state＝RAM 運行層）；閘門只讀 git 事實與 flow-state.json；主對話秘書連續承載所有段，技術證據不足時強制外部唯讀意見，主對話複核後自行承擔裁定；commit、判決、推進、路由、pass 出口由主對話獨佔。

Creator tools for making Agent Skills easier to understand and generated interfaces less generic.

Review evidence connecting product brands to legal entities and manufacturers, then produce traceable JSON and DOCX reports.

A multi-Skill learning system with Topic-local coaching, portfolio planning, read-only views, and a private GitHub-backed Learning Vault.

Provider-neutral, token-aware, evidence-driven SDLC harness for coding agents.

agentparty-channel-probe is an open-source Claude Code plugin indexed from GitHub. Review its manifest and source before installing.

Connect coding agents and humans through durable AgentParty channels, with safe launch guidance for Claude Cross-session coordination.

Strike the wrong agent behavior; Popper compiles surviving preferences into local Claude Code rules.

De-identify your AI coding-agent session logs and pack them into an archive you can hand to someone else. Reads local files, writes local files, makes no network calls.

pulp is an open-source Claude Code plugin indexed from GitHub. Review its manifest and source before installing.

pulp is an open-source Claude Code plugin indexed from GitHub. Review its manifest and source before installing.

Claude Code plugin for the Pulp audio framework — build, test, design, and ship audio plugins and apps

A 股分析 skill 包，包含 12 个 skill（8 核心 + 4 变体）：单股分析、大盘复盘、板块分析、持仓检查、选股策略、策略回测、财务分析、投资研究、帮助、学习（v1.17.0 移除 monitor，v1.20.0 起稳定）

Source-grounded RuvNet knowledge, lifecycle enforcement, and learning for Codex.

A plugin that automatically accumulates ADRs, feature summaries, and TODOs in an Obsidian Vault, per development project

Turn long videos into captioned short clips, schedule social posts, and manage channel automations from ChatGPT and Codex.

Make your coding agent write modern Godot 4.7+ best practices — lightweight and token-efficient.

Plugin d'exemple servant de gabarit pour les futurs plugins.

AI Agentic Workflow for Claude Code: nineteen atomic, stack-agnostic skills + subagents that carry a feature from idea → spec → design → data-model → API → tasks → tests → implement → review → ship, plus a TDD implementation engine and a living roadmap. AI is the first lens (design through it, not bolt it on); the engineer stays architect and quality controller (every stage is Socratic and gated, not a wall of output); a domain anchor grounds the work (survey maps the repo before anything else); verification is a loop, not an afterthought (plan-tests maps every acceptance criterion to a test, implement is TDD, review runs in clean context so whoever wrote the code doesn't approve it alone); and honesty about uncertainty is built in (clarify runs a devil's-advocate pass before code exists). Start with survey, walk the gated backbone — each stage reads the previous artifact and hard-refuses if missing — keep the Now/Next/Later roadmap in sync, and codify what you learn back into the shared library.

Run a Codex CLI review as a fresh, one-shot process (generic codex exec with a hand-built prompt) instead of a shared long-lived app-server session. Bundles a standalone /codex-review command.

Run /clear-prep before /clear to back up durable memory and in-progress session state; a bundled hook then auto-restores the handoff note right after.

Shell-level guardrails for Claude Code. Command blocking, code security scanner, self-teaching, token economy, agent routing, MCP profiles, and session memory.

A skeptical QA agent with memory: baseline → delta runs (NEW/REGRESSED/RESOLVED), flaky quarantine with expiry, and evidence-cited release verdicts you can defend.

An agentic development lifecycle (ADLC) for Claude Code: intents before code, feature/bug-fix/refactor workflows with checkpoints and hook-enforced gates, parallel work in worktrees, an independent reviewer with autonomy tiers and a maintain loop for GitHub Actions and GitLab CI, traces and cost per run, evals and a learning loop. Technology skills come from the adk-backend, adk-desktop and adk-architecture plugins in the same marketplace (agentic-development-kit). Quiet in any project that has not opted in.

10 workflows couvrant 29 intentions de rénovation et d'entreprise BTP avec les actions du MCP RenooLab.

A deliberately hostile plugin used only by the scanner tests

An ordinary plugin that must not trip a single rule

Deployment automation. Includes 1 skill and 2 MCP servers.

Database tooling for agents. Includes 1 skill and 2 MCP servers.