Recently added
Plugins grouped by the day they were added to the directory, newest first.
Sep 13, 2026
- multi-model-orchestrator0stars
Route software tasks across current Claude, GPT-5.6/GPT-6, and Grok 4.7 models with task-sized reasoning and independent review
CodexClaude Code5 skills - merge-guard0stars
Post-merge verification tail: diffs merged main against the pre-merge state to catch junk files leaked by squash merges and unintended path changes, checks configurable grep-based business invariants (attribution parameters, forbidden patterns), auto-opens a cleanup PR for leaked junk, and guides an adjacent-behavior regression spot-check. Complements pre-merge review gates.
CodexClaude Code1 skill - irreversible-guard0stars
PreToolUse gate that blocks Bash commands with no practical local undo — de-obfuscates ssh/docker-exec transports, heredocs, command chains, and env prefixes, then matches a tiered deny-set (irreversible-everywhere ops block unconditionally; locally-reversible ops block only when the command names production)
CodexClaude Code - i18n-parity0stars
Vendorable zero-dependency translation key-parity gate for multilingual next-intl apps — diffs locale catalogs (presence, empty values, ICU args, leaf shape, namespaces) with declarable waivers, wired as a hard CI/pre-push gate
CodexClaude Code2 skills - google-ads-strategist0stars
Senior Google Ads strategist — designs campaigns through iterative improvement, verifies via Chrome (Ad Preview Tool, SERP, Transparency Center), and creates campaigns in Google Ads UI in PAUSED state for investor review. Enforces single-variable hypothesis discipline, buyer-intent + product-value gates, and accumulates per-advertiser learnings. Chrome-first, no API dev-token needed.
CodexClaude Code21 skills - gh-issue-fetch0stars
Fetch GitHub issue details with auth-gated images downloaded locally, plus epic task-list resolution — so Claude can read issue screenshots and reason about epic progress
CodexClaude Code1 skill - gemini-cli0stars
Integrate Google's Gemini CLI into Claude Code for second opinions, dual code reviews, and AI-assisted explanations
CodexClaude Code5 skills - fleet-propagate0stars
Idempotent fleet-wide config propagation: enumerates host, dev containers, init scripts, and container-creator skills from a per-host manifest, applies changes as marker-delimited managed blocks (rerun-safe), verifies each target, bakes changes into creator skills so future containers inherit them, and reports a per-target changed/verified/needs-manual matrix.
CodexClaude Code1 skill - emails-to-github-issues0stars
Turn emails from named senders into GitHub issues — IMAP fetch via Proton Bridge, HTML-body extraction, thread grouping that respects customer-side numbering, inline-screenshot handling, and scope confirmation before writes
CodexClaude Code1 skill - comfyui0stars
Manage ComfyUI workflows, models, and API — build workflows from natural language, queue executions, monitor progress, and download models
CodexClaude Code7 skills - codex-subagent0stars
Drive GPT-6 Astra as scoped subagents with task-routed low-xhigh effort and explicit max/ultra
CodexClaude Code4 skills - browser-test-router0stars
Multi-model delegation for browser testing — Kimi K2.5 (via opencode) for execution, Opus for analysis
CodexClaude Code2 skills - analyst-companion0stars
Real-time in-person meeting analyst — live transcription with speaker diarization via self-hosted WhisperLiveKit, surfaces live clarifying questions on a Tailscale-only console, answers voice commands using the on-screen Chrome context, and creates reviewed Plane work items at meeting end
CodexClaude Code3 skills - agent-sync0stars
Mirror AGENTS.md to CLAUDE.md for Codex-first projects. In Codex, AGENTS.md is the source of truth; Claude Code keeps its existing Claude-to-AGENTS generation behavior.
CodexClaude Code4 skills - statiker0stars
Certify the design bears load before construction: a free investigation/design loop held by five forcing points — recorded decisions, fresh-context attack, dispatchable-design stop rule, no-design implementation, isolated executed verify.
Claude Code1 skill - chitta4stars
Persistent memory and code intelligence for Codex. Learns preferences, remembers codebases, anticipates needs. 150+ tools. Local-first.
Codex30 skills1 MCP server - chitta4stars
Persistent memory and code indexing for Claude Code and Codex through a local daemon, 344 documented tool entries, and session hooks. OpenCode is an optional review backend through chitta-bridge.
Claude Code30 skills - weproxa0stars
Inspect HTTP traffic and build API mocks with WePROXA's local MCP server and debugging skills.
CodexClaude CodeAgent Plugins3 skills1 MCP server - research-reader0stars
Build concept-organized research EPUBs from documentation sites, GitHub repositories, papers and blog posts: source inventory, license checks, full-text chapters for openly licensed material, research-entry summaries for everything else, and validation. Uses the research-reader CLI.
Claude Code1 skill - professor-mind-reader0stars
Reverse-engineer a marking rubric into the handful of criteria that actually move the mark, then audit a draft against them.
Claude Code1 skill1 MCP server Predict where a render, simulation or training job will bottleneck, and what to change first.
Claude Code1 skill1 MCP server- podcast-video-studio0stars
Turn a long recording into a scored cut list of clips that stand alone, with destinations, titles and hooks matched to each segment.
Claude Code2 skills1 MCP server Learn a designer's repetitive command patterns and surface the next likely action, cutting the hotkey tax out of a creative workflow.
Claude Code1 skill1 MCP server- mental-health-chatbot0stars
Structured non-clinical check-ins with escalation always on: routes to real resources, never diagnoses, never handles crisis, and produces supervisor summaries and an audit log.
Claude Code2 skills1 MCP server - haptic-feedback-mapper0stars
Eliminates context-switching: quantifies the cognitive load of checking a screen, maps status checks to a small haptic vocabulary, and builds the Deep-Work Protection case in billable hours.
Claude Code1 skill1 MCP server - ghost-post-preview0stars
Read a draft post the way its audience and its ranking system will: fold test, hook audit, platform mechanics and a stated-assumption persona read, ending in the single most likely reason it underperforms — never a fabricated engagement number.
Claude Code3 skills1 MCP server - generative-digital-twin0stars
Build a governed style profile from your own past work so a generative apprentice produces drafts in your voice, not a generic one.
Claude Code2 skills1 MCP server - five-minute-fluency0stars
Turn a described gameplay problem into a one-page, personalised cheat sheet you can absorb before the next match.
Claude Code1 skill1 MCP server Map the modelled emotional arc of a cut against the attention its form expects — a structural reading of where a film is likely to lose people, never a measurement of any audience.
Claude Code1 skill1 MCP server- diagnose-by-sound0stars
Turn a described vehicle noise into a ranked, testable differential diagnosis with a safety verdict and a confirmation plan.
Claude Code2 skills1 MCP server - customer-sales-support0stars
AI-powered customer support agent trained on the company's own manuals and FAQs — pre-sales questions and support tickets alike: audits the help centre, restructures it so the agent answers without hallucinating, reports coverage gaps, and defines which questions must reach a person.
Claude Code2 skills1 MCP server Translate between generative code and the geometry it produces, keeping output assets lean instead of bloated.
Claude Code1 skill1 MCP server- basecamp-split0stars
Plan shared gear, food, weight and cost across a group trip so nothing is carried twice and nothing is forgotten.
Claude Code1 skill1 MCP server - refactor-python0stars
Deterministic, verified decomposition of oversized Python modules using rope + AST oracles, with fan-out/fan-in subagents in git worktrees and file-size guardrails.
Claude Code2 skills - workgraph0stars
Task coordination instructions for Claude Code, delivered through native session and subagent hooks.
Claude Code2 skills - goodboy0stars
Good Boy — 세상에 나쁜 AI는 없다(There Are No Bad Agents). agentic LLM-CLI 가드레일(인용 환각 검증·테스트 강제·축별 리뷰·도구 위임·재읽기 방지·컨텍스트 관리·재귀 폭주 방어·토큰 낭비 억제)
Claude Code8 skills - linkml48stars
Author, validate, review and release LinkML schemas with the linkml-scala CLI. Covers schema authoring, bootstrapping from RDFS/OWL/SHACL/JSON Schema/XSD or sample data, modelling-quality review, instance-data validation, and GitHub Actions setup. Auto-load for any task involving a LinkML schema.
CodexClaude Code1 skill - jarl0stars
Skill that turns your coding agent into the director of a feature branch: a committed issue loop where everything seen becomes an issue, a worker per issue works in its own worktree, evidence decides what merges, and the loop leaves no trace on main.
CodexClaude Code1 skill - nature-skills0stars
A growing collection of Claude skills for producing academic work at Nature-journal standard. Covers scientific figures (nature-figure), manuscript prose polishing (nature-polishing), citation retrieval and export (nature-citation), data availability statements and FAIR metadata (nature-data), and paper-to-PPTX presentation conversion (nature-paper2ppt). Future releases planned: statistical reporting, peer-review responses, methods writing, cover letters, and review articles. All rules derived from primary sources — published Nature papers, journal author guidelines, and structured writing curricula.
Claude Code9 skills - nereus-game2stars
게임 개발 하네스: 로블록스·Unity(폰)·Switch 어댑터, 게임 도메인 스킬. nereus 코어의 워크플로 위에서 동작한다
Claude Code12 skills - graphloops1stars
収束ループのグラフ実行版。convergence-loops の 4 ループを「節＋依存＋周回条件」の JSON に写し、機械（engine）が盤面・波・遮断・件数・連続カウント・上限を持ち、回す側の LLM は渡された節を実行して返すだけにする。実行版は /research-graph（/research-loop の実行版）と /review-graph（/review-loop の実行版）の 2 本。既存のコマンドは残し、本体は隣に置く。
Claude Code - claude-code-plugin-hud3stars
A compact statusline for Claude Code, written in Rust. Shows model, context window, cost, git, and rate limits.
Claude Code - solai0stars
Bootstraps, audits and adopts Obsidian vaults. Card classes are declared once in TOML and projected into the data dictionary, CLAUDE.md, the card skill, the Base view and the validator. Greenfield and retrofit. Declares its own failure mode: authoring-instead-of-shipping.
Claude Code5 skills - auditfellow0stars
Your internal audit methodology, loaded before the model writes: findings, risks, controls, testing procedures and workpapers, meeting notes, audit memos, data strategy and AI usage documents, GenAI application audits, in your team's format. Sign in with your AuditFellow account.
Claude Code1 skill1 MCP server - spine-aiml-docs1stars
Справка по самому продукту Spine AI/ML Edition: скилл check-spine-aiml-docs отвечает на вопросы о харнессе (бинарь arch-ml, доменные инструменты, fitness-правила, модельная матрица, мост библиотеки) по документации репозитория, а не по памяти.
Agent Plugins - rust1stars
Каноны идиоматичного, производительного и безопасного Rust для харнесса arch-ml: шесть навыков по первоисточникам (The Rust Book, Rust API Guidelines, Async Book и документация Tokio, The Rust Performance Book, Rustonomicon, Cargo Book) — идиоматичный код и API-дизайн, обработка ошибок, ownership, асинхронность и многопоточность, тестирование (unit/integration/doctests, proptest, insta, criterion, Miri), структура проекта, Cargo, workspace, CI и semver, профилирование и оптимизация, unsafe и FFI с SAFETY-контрактами.
Agent Plugins6 skills - patterns-resilience1stars
Паттерны устойчивости распределённых систем по каталогу Azure Cloud Design Patterns (Well-Architected): circuit breaker + retry, bulkhead, queue-based load leveling, cache-aside, rate limiting/throttling, health endpoint monitoring. Дистилляты с силами, trade-offs, композициями и банковскими примерами; источник указан в каждом скилле.
Agent Plugins1 skill - patterns-integration1stars
Паттерны интеграции и распределённых данных из pattern language Криса Ричардсона (microservices.io): saga, transactional outbox, CQRS/API composition, strangler + anti-corruption layer, idempotent consumer. Дистилляты с проблемой/силами/вариантами/антипаттернами и банковскими примерами; источник указан в каждом скилле.
Agent Plugins1 skill - aws-builders-library1stars
Дистилляты Amazon Builders' Library — инженерные практики распределённых систем от команд AWS (Brooker, Yanacek, Gabrielson, Weiss, Magerramov): таймауты/ретраи/джиттер, отказ от fallback, load shedding, статическая стабильность, выборы лидера, 8 мод отказа, очереди и бэклоги, fairness в мультитенантности. Каждый скилл — с банковским маппингом и URL первоисточника.
Agent Plugins7 skills - aws-agentic-ai1stars
Агентные AI-паттерны AWS Prescriptive Guidance (гайд 'Agentic AI patterns and workflows on AWS', 2026): таксономия из 10 паттернов агентов, LLM-workflows (chaining, routing, parallelization, orchestration, evaluator loops), агентные workflow-паттерны (saga orchestration, supervisor, scatter-gather) — с маппингом на банковский контур и сервисы AWS.
Agent Plugins1 skill