claude-code plugins

by Paulo Soares

Produção de criativos para redes sociais com orquestração de APIs de geração no centro — Ideogram, Leonardo, Flux, Runway e Freepik/Magnific via wrappers com trava de gasto; triagem de custo, engenharia de prompt, consistência visual e curadoria de variações. Figma como camada opcional de composição. 19 skills, 4 subagentes, revisão de acessibilidade obrigatória.

by Paulo Soares

Configure and manage Claude Code status line — installs a rich, configurable bash status bar (single-jq-pass, dynamic colors, formatted tokens, truncated paths) showing model, git branch, context usage, rate limits, and token counts.

by Paulo Soares

Stack-agnostic UI/UX engineering framework for Claude Code — 6 real subagents + inline orchestrator/theme personas, 5-phase workflow with human gates, Figma integration, WCAG 2.1 AA, deterministic SQLite audit via hooks. Supports shadcn-vue, shadcn/ui, PrimeVue, Untitled UI, Tailwind Plus, and custom stacks.

by Paulo Soares

Integração com YouTrack Cloud via MCP oficial embutido (2025.3+) — issues CRUD, busca com query language, comentários, tags, artigos da knowledge base e log de trabalho. Fallback REST autenticado para o que o MCP não cobre: agile boards, sprints, relatórios de tempo, anexos e operações em lote. Setup guiado com token fora do repositório.

by Tony Jiang

中文版:/send-to <会话名> 把消息转达给本机另一个 Claude Code 会话——会话名前缀/子串模糊匹配,多个或零个命中亮列表让用户挑,消息强制自包含,[ref] 确认照抄重发,发送后如实汇报挂起状态;跨 profile 窗口发现互不可见但 uds: 地址可直发(标准路径),配 SessionStart 注册 hook 把 socket 清单翻译成身份注册表(默认开,touch ~/.claude/.cc-session-registry-off 全局停用),四级找目标阶梯,文件交接降级为最后兜底;HAPI 会话里 ping_peer 可用但绝不传 force(被拒即 force 会杀掉对方 Claude 进程),被拒改走 inspect_peer 的 claudeSessionId → 注册表 → SendMessage;经 ping_peer 发的消息须自报家门。基于跨会话消息功能(Claude Code v2.1.224+,macOS/Linux)

by Tony Jiang

English version: /send-to <session-name> relays a message to another Claude Code session on this machine — fuzzy prefix/substring name matching, shows the session list on zero/multiple hits, enforces self-contained messages, verbatim [ref] retry, honest held-delivery reporting; cross-profile windows are mutually invisible but reachable via a direct uds: address (the standard path), a SessionStart registration hook translates the socket list into a shared identity registry (on by default; touch ~/.claude/.cc-session-registry-off to disable globally), a four-tier target-finding ladder, file handoff as the final fallback; inside HAPI sessions ping_peer may be used but never with force (a refusal means force would kill the target's Claude process) — on refusal go inspect_peer's claudeSessionId → registry → SendMessage; messages sent via ping_peer must identify the sending window. Built on cross-session messaging (Claude Code v2.1.224+, macOS/Linux)

by Ālo Labs

Sidekick — AI coding-agent delegation workflows for Codex. Kay and Codex are the supported sidekicks on Claude Code, Codex, and Cursor hosts. Codex consumes the canonical Sidekick skills directly.

by Maxwell Volz

slashwork earner. /earn init authenticates and scaffolds an earner agent folder with run settings (model, bypass_permissions, default_duration); a bare /earn runs the default duration, /earn <goal> holds the live task feed over SSE, claims offloaded subagent tasks the moment they land, runs each with the folder's configured agent, and a SubagentStop hook submits the artifact with token usage. Earn credits and a per-class score for accepted work.

by Maxwell Volz

slashwork offloader. Interception is on by default: /work init authenticates (a bare /work with no token does the same), and from then on a PreToolUse hook routes self-contained subagent spawns to the offload network and returns the artifact in place, falling back to the local spawn on any miss. /work off pauses routing per project, /work on resumes it; a bare /work shows status and the dashboard.

by Tony Jiang

中文版:让 Claude 说人话、会提问——提问前自检清单(先核实/交代背景/黑话带解释/选项留组合位) + 表达纪律(S1~S6,含 S6 更新日志式汇报:多项进展或有未做/问题时用「本次完成/未完成/已知问题」三段体),规则源自作者 548 次真实 AskUserQuestion 抉择记录的数据挖掘。touch ~/.claude/.speak-human-always 即常驻

by Tony Jiang

English version: make Claude speak human and ask well — a pre-question self-check (verify first / brief the context / gloss the jargon / leave room to combine options) plus expression discipline (S1–S6, incl. S6 changelog-style reporting: multi-item progress or anything undone/wrong gets a Done/Not done/Known issues report), mined from 548 real AskUserQuestion decision records. touch ~/.claude/.speak-human-always to keep it always on

by Hermes Labs

Bounded public search evidence with source-explicit supersearch.search.v1 receipts (pinned hermes-supersearch 0.11.0); evidence, not answers

by Tony Jiang

中文版:SOP 截图前/改完一批 UI 后,把全站可交互元素系统性点一遍——agent-browser(Rust CLI,自带 Chrome,a11y 快照)驱动 + 通用化遍历引擎(config 外置、默认拦截名单只增不减、八分类记账 + 可选孤儿功能对账:代码清单 vs 遍历实际到达,查「功能实现了但前端没入口/入口打不通」)。安全边界:破坏性按钮只记不点、confirm/prompt 一律驳回、出域双层防护(--allowed-domains 限显式导航 + 引擎逐击域名校验,出域即拉回),只用于自家或已获授权的站点。

by Tony Jiang

English version: before SOP screenshots or after a batch of UI changes, systematically click every interactive element on the site — driven by agent-browser (Rust CLI, bundled Chrome, a11y snapshots) plus a generalized traversal engine (externalized config, an append-only default denylist, eight-category ledger, plus an optional orphan-feature audit: code inventory vs what the sweep actually reached, finding features with no UI entry point or a broken one). Safety boundary: destructive buttons are logged, never clicked; confirm/prompt dialogs are always dismissed; two-layer domain guard (--allowed-domains for explicit navigation + per-click hostname check, off-domain clicks pulled back on the spot); only for sites you own or are authorized to test.

by Tony Jiang

文档驱动的多代理开发工作流（指挥/执行/评审三层）：/scaffold 在项目里铺设 CLAUDE.md + docs 八件套脚手架，/whats-next 读文档判断下一步该干什么，sop-generate 为已部署的 Web 应用生成带截图的中文业务 SOP；新会话自动认出「要做新项目」并静默铺设脚手架（auto-scaffold，可 opt-out）；docs-capture 三层 hook 自动把问答框拍板存入 DECISIONS.inbox、按信号词提醒决策/需求未落档、commit 前门禁催消化，防止关键决策与需求随对话蒸发；方法论含目标台账（需求当轮落账防蒸发）、四点评审纪律（链式对抗评审／评审报差异／安全加固措辞／生产红线前置到实现侧）、调研内部先行与组件索引收尾登记

by Tony Jiang

Doc-driven multi-agent development workflow (Conductor/Executor/Reviewer three tiers): /scaffold lays down CLAUDE.md + an eight-doc set in place; /whats-next reads the docs and tells you what to do next; sop-generate generates a screenshot-backed business SOP for an already deployed web app; new sessions auto-recognize "wants to build a new project" and silently scaffold it (auto-scaffold, opt-out available); docs-capture — three hook layers auto-log AskUserQuestion decisions to DECISIONS.inbox, nudge on decision/requirement signal words, and gate commits until the inbox is triaged, so key decisions and requirements don't evaporate; the methodology carries a goal ledger (log requirements the same round so they never evaporate), four review disciplines (chained adversarial review / delta-only reports / neutral hardening wording / production red lines in the implementer prompt), internal-first research and component-index registration at wrap-up

by yu

Design-first code development workflow for Codex and Claude Code with orchestrated implementation, testing, and review workers.

by NotGabriels

Fifty hand-tuned dark Claude Code themes with semantic status colors.

by sumik5

AI/LLM/エージェント開発スキル群を提供するプラグイン。6個のスキル（GenAIデザインパターン・AIエージェント構築・Web AI統合・promptfoo評価/レッドチーミング・AI支援開発方法論・AI開発セキュリティ戦略）を含む。

by Altug Sogutoglu

Auto-inject ArcadeDB graph context per project; slash commands for decisions, queries, indexing, and status. Pre-bundled hooks, zero runtime npm deps.

by blackhole contributors

Agent-agnostic backlog campaign orchestrator to empty the forge backlog.

by sumik5

資格・検定の学習支援プラグイン。2個のスキル（creating-flashcards: EPUB/PDF/スキャン本、または certificate-exam サイトの URL（kentei-lab.com・studying/member.studying.jp・Whizlabs・shikaku-drill.com）からAnki MCP Server経由でフラッシュカードを一括作成。URLはホスト名で自動判定し内蔵の収集ロジックで問題文・正解・解説を取得してからAnki登録する・converting-content: 画像ベースEPUB→テキストOCR変換とLM Studioによるローカル翻訳）と1個のコマンド（improve-creating-flashcards: creating-flashcardsの自己改善）を含む。

by Robert Pauls

Bounded, value-aligned self-improvement workflows (ESRA) for Claude Code: five on-demand skills plus a dependency-free local runtime for evidence logs, baselines, experiments, audits, and human-reviewed oversight proposals.

by Christopher Beaulieu

Claude Code efficiency and hygiene toolkit. Surfaces token spend across the 5h / 7d / Sonnet-7d billing windows with per-model / per-agent / per-skill attribution, and audits your effective Claude Code configuration (custom + plugin agents and skills) for overlap and conflicts. Ships an interactive dashboard, a conversational usage-analysis skill, and a structured config-audit skill.