agentsclimarketplace

Context & AI engineering

25,729 rows, by stacks then stars

1,865 of these skill files have been read, by 1,328 distinct authors, and what they tell an agent counted. What Context & AI engineering authors agree on, and what they forbid

  • Mem

    wjb127/claude-skills/claude-mem/skills/mem Skill

    0★ repo

    Per-project external memory. Splits work notes into topic-based markdown files under the current project's memory/ folder. Saves context window by recalling only the needed chapter. Subcommands: save, recall, update, split, index, catchup, visualize. Separate from Claude Code's built-in /memory (conversation notes) — this is for code/deployment/decision records. 프로젝트별 외장 기억소. 현재 프로젝트의 memory/ 폴더에 주제별 md 파일로 작업 내용을 분할 저장. 컨텍스트 윈도우 절약 + 필요한 챕터만 호출하는 패턴. 서브커맨드: save, recall, update, split, index, catchup, visualize. Claude Code 기본 /memory (대화 메모) 와는 별개 — 이건 코드/배포/결정 기록용. Triggers (Korean): "메모리에 저장", "기억해줘", "memory 정리", "어디까지 했지". Triggers (English): "save to memory", "remember this", "where were we", "memory dump".

  • Codex implementation

    Studio-Moser/skills-n-stuff/plugins/pm/skills/codex-implementation Skill

    0★ repo

    Ask the Codex CLI (running whatever model it's configured with) to implement a bounded, well-specified change in the repo. This is how the Codex model is invoked for implementation work. Use when the model-selection rubric routes bulk or mechanical implementation to Codex, when the user asks Claude to have Codex implement something, or when a parallel implementation agent should produce a patch in a worktree. For work Claude should own directly, implement it yourself instead.

  • Compact io

    Evan-Daruwalla/claude-skill-suite/compact-io Skill

    0★ repo

    Always-active output compression: keep every response filler-free, jargon-free, and maximally dense so it spends fewer output tokens and bloats future context less. Apply to all interactions. Also handles explicit requests to make an answer shorter/denser/less jargony ("compress this answer", "make this denser", "plainer words"). Yields to project/task instructions on format and required detail; governs default density only. For compressing a prompt/doc/system-prompt FOR REUSE across sessions, that's token-squeeze's job (deterministic, guard-checked), not this skill.

  • Llm eval harness

    Evan-Daruwalla/claude-skill-suite/llm-eval-harness Skill

    0★ repo

    Measure how far a cheaper model falls from your flagship model's quality bar on your real task types, deterministically and reproducibly — no API key, no LLM-judge. Scores model outputs by deterministic CHECKS (format, discipline, no-fabrication, surgical-scope) and by line-SIMILARITY to captured flagship goldens, and appends every run to a ratchet so the gap is trackable over time. Use when the user says "eval the model", "measure the model gap", "run the eval harness", "score this output", or when deciding whether a cheaper model is good enough to switch to.

  • Skill vet

    Evan-Daruwalla/claude-skill-suite/skill-vet Skill

    0★ repo

    Evaluate an external Claude Code skill, plugin, or MCP server (usually a GitHub URL) before installing it. Use when the user says "look at this skill", "evaluate whether these skills would be useful", "install this skill", or pastes a repo/marketplace link. Produces a verdict (install / skip / install-modified) with reasoning, then installs cleanly on approval.

  • Token squeeze

    Evan-Daruwalla/claude-skill-suite/token-squeeze Skill

    0★ repo

    Deterministic, no-LLM compressor that rewrites verbose English into fewer tokens for text that will be REUSED — saved prompts, system prompts, skill bodies, docs pasted repeatedly. Use when the user asks to "compress this file/prompt/doc for reuse", "shrink this system prompt", "run token-squeeze", or wants to cut the token cost of a reusable text asset. Runs a bundled Node CLI (no API, no model call). NOT for compressing live chat turns — those tokens are already spent on arrival; this only pays when the compressed text replaces the original in future contexts. Meaning-leaning, not lossless: it substitutes verbose phrases for equivalents and protects code, URLs, paths, quotes, numbers, and negations.

  • Agents get started

    dkmqflx/claude-tools/.claude/skills/agents-get-started Skill

    no license0★ repo

    Use when a developer wants to create a new agent project or get started with AgentCore. Handles framework selection, project scaffolding, first deploy, and first invocation. Triggers on: "build an agent", "create an agent", "get started", "new project", "agentcore create", "which framework", "Strands vs LangGraph", "hello world agent", "first agent", "create MCP server", "host MCP server", "agentcore dev", "dev server", "what port", "local development". Not for adding capabilities to existing projects — use agents-build or agents-connect. Strands vs LangGraph in a migration context routes to agents-build, not here. Connecting to an existing MCP server routes to agents-connect, not here.

  • Amazon bedrock

    dkmqflx/claude-tools/.claude/skills/amazon-bedrock Skill

    no license0★ repo

    Builds generative AI applications on Amazon Bedrock. Covers model invocation (Converse API, InvokeModel), RAG with Knowledge Bases, Bedrock Agents, Guardrails, and AgentCore. Use when invoking models, setting up Knowledge Bases, creating agents, applying guardrails, deploying to AgentCore, troubleshooting Bedrock errors (ThrottlingException, AccessDeniedException), or choosing models (Claude, Llama, Nova, Titan). ALSO USE for prompt caching setup and debugging, quota health checks and throttling diagnosis, cost attribution and tracking, migrating between Claude model generations (4.5 to 4.6 to 4.7), chunking strategies, API selection (Converse vs InvokeModel), guardrail capabilities, and model selection. NOT for custom model training, Rekognition, or Comprehend.

  • Deep agents memory

    dkmqflx/claude-tools/.claude/skills/deep-agents-memory Skill

    no license0★ repo

    INVOKE THIS SKILL when your Deep Agent needs memory, persistence, or filesystem access. Covers StateBackend (ephemeral), StoreBackend (persistent), FilesystemMiddleware, and CompositeBackend for routing.

    The same file is in 1 other repository

  • Deep agents orchestration

    dkmqflx/claude-tools/.claude/skills/deep-agents-orchestration Skill

    no license0★ repo

    INVOKE THIS SKILL when using subagents, task planning, or human approval in Deep Agents. Covers SubAgentMiddleware, TodoList for planning, and HITL interrupts.

    The same file is in 1 other repository

  • Langchain rag

    dkmqflx/claude-tools/.claude/skills/langchain-rag Skill

    no license0★ repo

    INVOKE THIS SKILL when building ANY retrieval-augmented generation (RAG) system. Covers document loaders, RecursiveCharacterTextSplitter, embeddings (OpenAI), and vector stores (Chroma, FAISS, Pinecone).

  • Managed deep agents

    dkmqflx/claude-tools/.claude/skills/managed-deep-agents Skill

    no license0★ repo

    INVOKE THIS SKILL when creating, running, or operating a Managed Deep Agent against the LangSmith /v1/deepagents private-preview REST API. Covers the agent → MCP server → thread → streamed run flow, tool/interrupt configuration, and the agent file tree (AGENTS.md, skills/, subagents/, tools.json).

  • Storing and querying vectors

    dkmqflx/claude-tools/.claude/skills/storing-and-querying-vectors Skill

    no license0★ repo

    Store and query vector embeddings using Amazon S3 Vectors, a cost-effective long-term vector storage service with its own API namespace (s3vectors). Triggers on: create S3 vector bucket, vector index, store embeddings, semantic search, RAG vector storage, similarity search, vector database, migrate from other vector databases. Do NOT use for: querying tabular data (use querying-data-lake), S3 object storage, or hundreds/thousands of sustained QPS (use OpenSearch).

  • Swarm

    dkmqflx/claude-tools/.claude/skills/swarm Skill

    no license0★ repo

    Dispatches many independent items in parallel: create a table, fan out to subagents, aggregate results. One row = one unit of work.

  • Worktree done

    dkmqflx/claude-tools/.claude/skills/worktree-done Skill

    no license0★ repo

    Remove the current git worktree, switch the main checkout to the base branch, and delete the feature branch. Use when the user has finished work started with the worktree skill (or `/worktree`) and wants to clean up — phrases like "워크트리 정리해줘", "워크트리 삭제해줘", "clean up this worktree", "remove worktree", "/worktree-done". Do NOT use from the main checkout (not a worktree) — that is the branch-done skill.

    The same file is in 1 other repository

  • Worktree

    dkmqflx/claude-tools/.claude/skills/worktree Skill

    no license0★ repo

    Create a new git worktree on a fresh branch and switch into it before starting work. Use when the user asks to start a task in a new worktree, work in isolation from the main checkout, or invokes "/worktree" — phrases like "워크트리 만들어줘", "새 브랜치 파서 작업해줘", "worktree에서 작업 시작". Do NOT use for removing or cleaning up an existing worktree.

    The same file is in 1 other repository

  • Repo insight skill

    sqliang/deep-insight-engines/engines/repo-insight/repo-insight-skill Skill

    0★ repo

    专门用于深度分析和解读本地或 GitHub 源代码仓库的 Agent Skill。在 Claude Code 中全自动运行,包含质量 Review、迭代优化闭环、并行处理和 subAgent 支持。当用户说"帮我分析这个 GitHub 仓库"或"分析当前目录的代码"时,使用此技能。

  • Skill creator

    cloverink/shipwright/skills/skill-creator Skill

    0★ repo

    Meta-skill that teaches how to create effective Claude Code skills. Covers anatomy, progressive disclosure, validation, and packaging.

  • Tap triggers

    LeonTing1010/taprun/plugins/tap/skills/tap-triggers Skill

    0★ repo

    Declare WHEN a saved tap runs unattended — ~/.tap/triggers/ trigger.json declarations compiled idempotently into launchd plists (dev.taprun.trigger.* namespace), zero LLM tokens per fire. Foreground-gated plans (trusted:true) are refused at declaration time. Use when the user wants a tap to run on a schedule / on file change ("run this tap every day", "check back on a schedule", "schedule this tap", "put this on launchd", "tap trigger"). NOT for Claude Code routines/cron (those wake an agent and burn tokens per fire) — this is the zero-token trigger layer for already-compiled plans.

  • Image prompt engineer

    prvthmpcypher/skills-design/skills/image-prompt-engineer Skill

    0★ repo

    Expert photography prompt engineer specializing in crafting detailed, evocative prompts for AI image generation. Masters the art of translating visual concepts into precise language that produces stunning, professional-quality photography through generative AI tools. Use when the user asks about image prompt engineer, needs this workflow, or requests related deliverables.

  • Debate

    AI-Driven-School/aiki/.claude/skills/debate Skill

    0★ repo

    Multi-Agent Debate (Solver/Proposer/Critic/Defender/Checker 5エージェント反証) で主張・実装・判断を独立検証。学習源の異なるAI (Claude/Gemini/Codex/Grok+WebSearch) を意図的に対立させてハルシネーション・事実誤認を検出。v1.2 で forecast モード追加: 将来予測系の主張には MiroFish 式の agent-based simulation (ステークホルダー・ペルソナ世界を多ラウンド) でシナリオ分布を生成。v1.3 で precision 強化: Defender(擁護者)を追加し攻撃3:擁護0の非対称を解消、かつ「反論=有罪」をやめ Checker が一次情報で裏取りできた反論のみを採用 (false positive 抑制)。`/debate <主張ファイル or テキスト>` で起動。MARCH/Microsoft CORE研究に基づく実装で false positive -25.8% 期待。

  • Rendercv resume handoff

    ArthurZakirov/ProofStack/skills/rendercv_resume_handoff Skill

    no license0★ repo

    Use when a user wants to turn Career Positioning OS resume content into a RenderCV YAML file, install or use the external RenderCV agent skill, or render a resume/CV PDF without reimplementing RenderCV's schema, CLI, themes, or PDF generation.

  • Setup mcp

    az9713/youtube-transcript-to-skills/.claude/skills/setup-mcp Skill

    no license0★ repo

    Find, install, and configure an MCP server for a specific service. Handles discovery, installation, and verification.

  • Code review chunk dispatch

    Ed3Design/ed3design-skill-bundles/code-quality/skills/code-review-chunk-dispatch Skill

    0★ repo

    Use when the code-review-backlog is large (>30 commits OR >5,000 LoC changed since last review) AND the caller has Agent/Task-tool dispatch capability (.e., top-level orchestrator or main Claude session). Trigger on phrases like "code-review backlog is big", "no review for weeks", "many commits without review", "review of 100+ commits", "how do I review these 200 commits", "chunk review", "parallel code-reviewer subagents". Do NOT load for single-PR review (use superpowers:requesting-code-review directly), for <30-commit backlog (single subagent suffices), for non-Git codebases (chunking-by-SHA-range presumes git workflow), or when running as a subagent yourself without Agent-tool access (the chunk-dispatch pattern is then unexecutable — use risk-based-triage instead, documented in the skill's Fallback section). Complements (sub-of) `superpowers:requesting-code-review` and `superpowers:dispatching-parallel-agents`.

  • Mcp server stdio to http migration

    Ed3Design/ed3design-skill-bundles/code-quality/skills/mcp-server-stdio-to-http-migration Skill

    0★ repo

    Use when an MCP server uses `stdio` transport AND is spawned per-sub-agent via `uvx --from git+https://...`, `npx git+https://...`, or an equivalent on-demand-installer, AND you observe ANY of: (a) `.in_use/` lock-directories accumulating as zombies, (b) 10+ second startup per sub-agent (uvx re-resolves the Git source), (c) one server process per sub-agent instead of a shared instance, (d) the server "hangs" at session start. Fix-pattern: migrate to `streamable-http`/`http` transport with a persistent server (macOS LaunchAgent or systemd-unit) on a fixed localhost port, and point the client config at the HTTP URL. Result: one shared process, no zombies, startup O(seconds)→O(milliseconds). Trigger on phrases like "MCP zombie processes", "MCP stdio slow startup", "uvx --from git is slow", "streamable-http LaunchAgent". Do NOT load for servers designed for short-lived stdio (file-system MCP), for non-Git pip-installed servers that already start fast, for first-time MCP setup, or for server-selection questions.

  • Subprocess ssh arg quoting via shlex

    Ed3Design/ed3design-skill-bundles/code-quality/skills/subprocess-ssh-arg-quoting-via-shlex Skill

    0★ repo

    Use BEFORE writing or debugging Python `subprocess.run(["ssh", host, "cmd", "arg1", "arg2"])` calls where remote args contain shell-metacharacters — pipes `|`, parens `()`, semicolons `;`, dollar-signs `$`, backticks, quotes, glob `*?`, tabs `\t`, or empty strings. The naive list passed after host is silently joined by SSH with spaces and re-evaluated by the REMOTE shell. Symptoms: 'syntax error: unexpected end of file', 'option requires an argument', 'command not found', or silently-wrong arg parsing. STOP and use `shlex.quote()` on EACH remote-side arg, then `' '.join(quoted_args)` to produce ONE string passed as a SINGLE post-host arg. Trigger on phrases like "ssh + docker exec doesn't work", "psql via SSH gives syntax error", "MCP server SSH tunnel", "Bash syntax error from remote", "tab character disappears in SSH". Do NOT use for `shell=True` calls, SSH config aliases without metacharacter-args, pure tunnel/SFTP, or pure-local subprocess.

  • Skill tdd promotion workflow

    Ed3Design/ed3design-skill-bundles/skill-system-meta/skills/skill-tdd-promotion-workflow Skill

    0★ repo

    Use when promoting an existing DRAFT/STUB skill (suffix `-DRAFT` or `-STUB` in `name:` field, OR description starts with "STUB —") to GA (auto-discoverable). Different from `superpowers:writing-skills` which covers CREATE-from-scratch — this skill covers the LIFECYCLE-STAGE from "skeleton-exists" to "production-ready". Requires Agent/Task-tool for parallel RED+GREEN subagent-Dispatch (Step 3) — if the caller does NOT have Agent-tool (e.g. running as a general-purpose-subagent themselves), STOP and report-up; do NOT silently fall back to name-rename-only. Trigger on phrases like "promote this DRAFT skill", "TDD for the remaining STUBs", "make it auto-discoverable", "skill promotion cycle", "release skill from -DRAFT", "TDD promotion cycle for N skills". Do NOT load for creating a new skill from scratch (use `superpowers:writing-skills`), for editing GA-skills (just edit), when no DRAFT/STUB-suffix exists (skill is already GA), or when target skill file does not exist (no work to do).

  • Subagent mode selection continuous vs review between

    Ed3Design/ed3design-skill-bundles/skill-system-meta/skills/subagent-mode-selection-continuous-vs-review-between Skill

    0★ repo

    Use when starting a multi-task workflow via `superpowers:subagent-driven-development` (SDD) and the user must choose between two execution modes: (A) Continuous Execution — all tasks dispatched back-to-back without checkpoint (SDD default), or (B) Review-Between-Tasks — user inspects each task's output before authorizing the next. Mechanical-implementation tasks (TDD with complete code in plan, deterministic outcomes) are continuous-safe. Judgment-heavy tasks (forensics, interpretation of numbers, strategic decisions, architectural pivots based on subagent output) require review-between because the next task's prompt depends on prior outcome interpretation. Trigger phrases like "should I run continuous or with pauses between tasks", "review between steps", "subagent-driven with checkpoints", "pause after each task". Do NOT load for single-task workflows, non-SDD workflows, purely mechanical TDD with complete code in plan, or the very first task of a workflow.

  • Subagent self reflection prompt pattern

    Ed3Design/ed3design-skill-bundles/skill-system-meta/skills/subagent-self-reflection-prompt-pattern Skill

    0★ repo

    Use when dispatching a subagent to test a skill, evaluate a tool, or perform any non-trivial task where YOU (the caller) want meta-feedback about the experience — what worked, what was unclear, what was missing. Include a mandatory "## Skill-Self-Reflection" section in the prompt asking 3-4 introspective questions. Without this, you only get task output, not improvement signal. Trigger on phrases like "dispatch subagent for skill test", "skill TDD pressure-test", "evaluate this tool via subagent", "subagent meta-feedback", "how do use subagents for skill improvement". Do NOT load for production-task dispatch (only output matters, not reflection), for performance-critical dispatches (reflection adds tokens), or when the caller already has a strong skill-content hypothesis to test against (then use direct A/B comparison instead).

  • Set goal

    robmcquade/claude-skills/set-goal Skill

    0★ repo

    Compose a strong condition for Claude Code's built-in /goal autonomous loop, then hand it back ready to paste. /goal keeps Claude working across turns until a fast evaluator model (Haiku by default) judges the condition met — and that evaluator sees only the main transcript, not your files and not a subagent's internal work unless Claude restates it, so the condition must be provable from what Claude surfaces. This skill turns a rough aim into a well-formed condition — the three things the docs call for (a measurable end state, a stated check, constraints that must hold) plus the turn or time cap the docs recommend separately — validates it against the transcript-only rule and the common failure modes, and explains the residual risks. User-invoked only. Adaptive — refines a draft you pass, or drafts from the project and asks only the gaps from a blank slate. Pass deep to adversarially stress the drafted condition with a fresh-context subagent.

  • Harness verify before move

    ClydeShen/harness-skill/skills/engineering/harness-verify-before-move Skill

    0★ repo

    Comprehensive pre-phase-transition sweep. Verifies phase exit criteria, syncs code status (git/lint/build/tests), audits and cleans design docs, cleans implementation-notes, prunes stale documents and memory, syncs GitHub project state (milestones, issues, status labels, board), updates README, writes a fresh memory snapshot, then outputs a clean project snapshot. Use when completing a phase, before advancing to next phase, or when user says "verify before move", "clean up before moving", "sync everything", "project snapshot", "clean slate", "phase transition". Typically consumes ~8K–15K tokens / <10% of a context window.

  • Session start

    ClydeShen/harness-skill/skills/engineering/session-start Skill

    0★ repo

    Briefs the current session by reading session state, handoff doc, and memory to determine the active phase, task, and context budget. Outputs a structured briefing and detects whether the phase should be skipped or reverted based on observable artifacts. Use at the start of every long-running agent session, or when user says "start session", "what are we working on", "resume", "pick up where we left off", "what's the active task".

  • Setup harness skills

    ClydeShen/harness-skill/skills/engineering/setup-harness-skills Skill

    0★ repo

    Creates CLAUDE.md/AGENTS.md from the harness template (if absent) and configures docs/agents/ so harness skills know the issue tracker, triage labels, domain docs, GitHub Project board, and session state location. Run before first use of triage, to-prd, to-issues, context-handover, or session-start — or if those skills appear to be missing context. Typically consumes <30% of a context window.

  • Skill cleanup

    ClydeShen/harness-skill/skills/productivity/skill-cleanup Skill

    0★ repo

    Audit and remove stale, renamed, or duplicate skills across all installed agent platforms (Claude Code, Kiro, Codex, Windsurf, Gemini, Pi, and more). Handles symlink and copy installs, detects artifacts, and generates OS-correct removal commands. Use when skills list feels cluttered, after renaming a skill, after uninstalling a plugin, or when user says "clean up skills", "remove old skill", "prune skills", "stale skill", "duplicate skill".

  • Reconcile claude md

    TomerAberbach/claude-config/skills/reconcile-claude-md Skill

    0★ repo

    🤖 My Claude Code config!

  • Memory

    tarangdeep-goel-by/claude-harness/claude/skills/memory Skill

    0★ repo

    Memory corpus quality rollup. /memory health reports capture/approval/utilization/hook-health; /memory review proposes harness tweaks.

  • Codebase memory

    ccronca/dotfiles/claude/skills/codebase-memory Skill

    0★ repo

    Use the codebase knowledge graph for structural code queries. Triggers on: explore the codebase, understand the architecture, what functions exist, show me the structure, who calls this function, what does X call, trace the call chain, find callers of, show dependencies, impact analysis, dead code, unused functions, high fan-out, refactor candidates, code quality audit, graph query syntax, Cypher query examples, edge types, how to use search_graph.

    The same file is in 1 other repository

  • Subagent driven development

    chenwei791129/agent-skills/hermes-skills/superpowers/subagent-driven-development Skill

    0★ repo

    Use when executing implementation plans with independent tasks in the current session

  • Find claude session

    chenwei791129/agent-skills/skills/find-claude-session Skill

    0★ repo

    Search across all past Claude Code sessions (across every working directory) to recall what was discussed and locate the original cwd. Use this skill whenever the user wants to find a past conversation, recall which directory a session was in, search their Claude history for a topic, or asks things like '我之前在哪個對話/目錄聊過 X', '幫我找之前討論 X 的 session', 'which directory did we talk about X in', 'search my Claude transcripts for Y'. Especially useful when the user opens throwaway conversations via `cd $(mktemp -d)` and later forgets which tmp directory hosted a discussion.

  • Grafana

    chenwei791129/agent-skills/skills/grafana Skill

    0★ repo

    Use when operating a Grafana instance from the terminal — listing/searching dashboards, querying datasources, inspecting alert rules, running PromQL/LogQL/TraceQL over metrics/logs/traces, or pulling/pushing dashboards as code (GitOps). Triggers on Grafana, gcx, PromQL, LogQL, dashboard, alerts/告警, 查 Grafana, dashboard 管理, 監控查詢. Use this instead of a Grafana MCP server or raw curl/grafana-backup/grizzly. Self-hosted focused.

  • Prompt engineering

    chenwei791129/agent-skills/skills/prompt-engineering Skill

    0★ repo

    Use this skill when you writing commands, hooks, skills for Agent, or prompts for sub agents or any other LLM interaction, including optimizing prompts, improving LLM outputs, or designing production prompt templates.

    The same file is in 24 other repositories

  • Agency deliverable handoff

    agency-shift/agency-shift-skills/agency-deliverable-handoff Skill

    0★ repo

    Package a finished deliverable for client handoff — produces the handoff message, the QA checklist that proves it's done, the asset link manifest, and the next-action ask. Use whenever the user says "ship this", "hand this over", "send to client", "deliverable is ready", or pastes a list of done items + asset URLs and asks for the handoff message. Do NOT use this skill for weekly status updates (use agency-weekly-report instead) or for internal team handoffs between two agency staff.

  • Prompt engineering

    Bbar0n234/llm-engineer-skills/skills/prompt-engineering Skill

    0★ repo

    Помогает писать системные промпты для LLM. Используй когда нужно написать prompt, system prompt, системный промпт, инструкцию для модели.

  • Session reflection

    gigayaya/gigachang-skills/skills/session-reflection Skill

    0★ repo

    Use when the user explicitly asks to reflect on, review, or learn from the current Claude Code session — e.g. "review this session", "reflect on this session", "why did that take so many tries", "what went wrong this session", or asks to turn this session's mistakes into project rules. Reads the session transcript, finds where Claude's output was rejected or required correction, distills the root causes, and proposes concrete project rules — integrating them into the project's existing rule system (CLAUDE.md, AGENTS.md, …) or proposing a simple one if none exists. Manual-trigger only — never auto-invoke; never write rule files without explicit user approval.

  • System pilot

    bogheorghiu/ex-cog-dev/makers-toolkit/skills/system-pilot Skill

    0★ repo

    What does 'done' look like here? Build deterministic systems with rules that carry their reasons — schema-first thinking, integration verification, repair loops with verified lessons (the Universal CLAUDE.md Protocol in command-with-reason register). Use for non-trivial projects or multi-agent system coordination. Not for typo fixes or exploratory conversation.

  • Agents md improver

    mjcreativelab/mjcreativelab-agent-prompts/skills/agents-md-improver Skill

    0★ repo

    CLAUDE.md / AGENTS.md / GEMINI.md などエージェント指示ファイル(プロジェクトメモリ)の監査・改善。 「CLAUDE.md を監査して」「AGENTS.md が最新か確認して」「指示ファイルを改善して」「プロジェクトメモリを最適化して」で起動。 リポジトリ内の指示ファイルを走査して品質レポートを提示し、ユーザー承認後に的を絞った更新を適用する。 コードベースの現状との整合を確認する定期メンテナンス向け(セッションで得た学びの取り込みは agents-md-revise を使う)。

  • Agents md revise

    mjcreativelab/mjcreativelab-agent-prompts/skills/agents-md-revise Skill

    0★ repo

    今回のセッションで得た学び(発見したコマンド・コードパターン・テストアプローチ・環境の癖・注意点)を CLAUDE.md / AGENTS.md / GEMINI.md などのエージェント指示ファイルへ反映する。 セッション終盤に /agents-md-revise で起動。 コードベース起点の定期監査・品質レポートは agents-md-improver を使う。

  • Claude code update review

    mjcreativelab/mjcreativelab-agent-prompts/skills/claude-code-update-review Skill

    0★ repo

    Claude Code のバージョンアップ後に、最新の公式推奨手法と現在の構成 (settings.json / commands / rules / skills / CLAUDE.md 等)を照合して改善提案を行う。 「Claude Code の最新機能」「アップデートレビュー」「設定見直し」「ベストプラクティス確認」 「/claude-code-update-review」で起動。 単一スキルの品質改善 (skill-improver) や PR/差分レビュー (smart-review) とは別物。

  • Empirical prompt tuning

    mjcreativelab/mjcreativelab-agent-prompts/skills/empirical-prompt-tuning Skill

    0★ repo

    agent 向けテキスト指示(skill / slash command / タスクプロンプト / CLAUDE.md 節 / コード生成プロンプト)の「指示の曖昧さ」を、新規 subagent(バイアスを排した実行者)に dispatch して実際に動かしてもらい、両面(実行者の自己申告 + 指示側メトリクス)で 評価して反復改善する手法。改善が頭打ちになるまで回す。 プロンプトや skill を新規作成・大幅改訂した直後、またはエージェントの挙動が 期待通りにならない原因を指示側の曖昧さに求めたいときに使う。 TODO 残留・リンク切れ・bash 構文など機械的な静的チェックが目的なら skill-improver を使う(本スキルは静的チェックは行わない)。

  • Memory dream

    mjcreativelab/mjcreativelab-agent-prompts/skills/memory-dream Skill

    0★ repo

    エージェントの記憶階層(MEMORY.md・notes・projects 等)を再編し、重複・矛盾・陳腐化を除去する consolidation(Anthropic Dreams の手動・git ベース再現)。 大規模リファクタ直後や 20〜30 セッション蓄積で記憶がノイズ化したとき、 「記憶を整理して」「dream して」「/memory-dream」で起動。 --codex-review-loop(-cdxrl)を付けると採用前に Codex レビューループを実施する (Claude Code + Codex プラグイン環境前提)。 --claude-review-loop(-cldrl)はコンテキスト隔離した Sonnet(effort max)レビュワーエージェントによる レビューループを実施する(Codex 不要・Claude Code の Workflow ツール前提。利用できない環境はレビュー未実施で degrade)。

  • Skill improver

    mjcreativelab/mjcreativelab-agent-prompts/skills/skill-improver Skill

    0★ repo

    既存の Claude Code スキル(`<skill-dir>/SKILL.md` + `references/` / `assets/` の一式。 仕様は本プラグイン README の「対象とするスキル定義」を参照)を、 skill-creator eval 委譲 + 公式ベストプラクティス差分 + 機械的静的チェックで仕上げるスキル。 TODO / FIXME 残留、参照ファイルのリンク切れ、.sh 構文(bash -n)、 コンテキスト管理設計(再 Read ルール・中間データ外部化・API レスポンス絞り込み)を Grep / Glob / bash で検証し、最新の Claude Code 公式推奨(frontmatter / SKILL.md 規約 / 動的コンテキスト注入)との差分を取って改善提案を A/B/C カテゴリで出力する。 「スキルを改善して」「品質チェック」「eval 回して」「500 行超えそう」 「SKILL.md を見直して」「スキルが効かない」「トリガーされない」「行数オーバー」 「最新のスキル規約に合わせて」等で起動。 指示の曖昧さを新規 subagent の実行実測で炙り出したいときは empirical-prompt-tuning を使う (本スキルは subagent dispatch は行わない)。 Claude Code 環境全体のレビューは claude-code-update-review、 PR / 差分レビューは smart-review、新規スキル作成は skill-creator を使う。

  • Gmail helper

    chienchuanw/chuan-skills/plugins/daily/skills/gmail-helper Skill

    no license0★ repo

    Act as the user's personal email secretary across all linked Gmail accounts (work, personal, semi, lighting, foufa) via the gmail-mcp server. Triage and label today's incoming mail, auto-applying existing labels on confident matches and proposing new labels with AskUserQuestion when nothing fits; then chase the status of older threads (awaiting reply, needs action, snoozed) and write the whole picture into the user's Obsidian daily journal. Use this skill whenever the user asks to "check my email", "do my inbox", "triage my mail", "what needs my attention", "go through my Gmail", "be my secretary", "any important emails", "follow up on old threads", or otherwise wants their multi-account inbox organized, labeled, or summarized into their journal — even if they don't say the word "skill" or name a specific account.

  • Apply coding rules

    chienchuanw/chuan-skills/plugins/dev/skills/apply-coding-rules Skill

    no license0★ repo

    Stamp a managed coding-discipline rules block into a project's CLAUDE.md (or AGENTS.md / GEMINI.md) — the four rules: think before coding, simplicity first, surgical changes, goal-driven execution, plus a guard on irreversible/git-history actions. This skill is the single source of truth for those rules; running it writes or updates the rules between managed markers, idempotently, leaving the rest of the file untouched. Use when the user wants to add, update, sync, or propagate the shared coding rules / coding standards into a repository — "apply the coding rules", "add the four rules to this project's CLAUDE.md", "sync our coding standards here", "stamp the coding discipline block" — even if they don't name the skill.

  • Connect

    chienchuanw/chuan-skills/plugins/gma2/skills/connect Skill

    no license0★ repo

    Install, register, and verify the gma2 MCP server so Claude can drive a grandMA2 lighting console (or onPC) over Telnet. Use this whenever the user wants to "connect to grandMA2", "set up the MA2 console", "install gma2 mcp", "hook up the lighting desk", asks why the gma2 tools aren't available, or is about to run any other gma2 skill (presets / setlist / bpm) and the server isn't connected yet. This is the entry point for the whole gma2 plugin — run it first.

  • Presets

    chienchuanw/chuan-skills/plugins/gma2/skills/presets Skill

    no license0★ repo

    Build grandMA2 Gobo / Color / Beam / Focus preset palettes from fixture-type XML profiles, programming the correct per-fixture-type DMX value for each named function so one logical preset works across mixed fixtures. Use whenever the user wants to "build presets", "make a color palette", "set up gobo/beam/focus presets", "program a preset pool", or hands over fixture XML files and a console to populate presets on. Requires the gma2 MCP server connected (run the connect skill first). Prefer this over hand-writing preset commands — it resolves fixture-specific values from the profiles instead of guessing.

  • Skill benchmark

    chienchuanw/chuan-skills/plugins/skill-optimize/skills/skill-benchmark Skill

    no license0★ repo

    Benchmark a skill's quality by analyzing its structure, description, documentation, and live value-add from multiple perspectives using parallel agents. Use when the user asks to benchmark, evaluate, score, audit, assess, or rate a skill's quality, or wants to know if a skill is worth keeping vs. deprecating. Also trigger when the user asks "how good is this skill", "is this skill useful", "should I keep this skill", "review skill quality", or wants a thorough multi-dimensional analysis of a skill before deciding whether to improve or remove it. Scoped to one skill: for a whole-repository health/security audit use health-audit, not this. Takes a skill name as argument (e.g., /skill-benchmark readme). Produces a scored report with actionable improvement suggestions. Do NOT use for creating new skills, editing skill code, or capturing gotchas — those are different workflows.

  • Sandboxed claude

    bensimpson-ch/claude-skills/sandboxed-claude Skill

    0★ repo

    Create a Podman dev container that runs Claude Code inside a sandboxed environment. Use when setting up a containerized development environment for any project, when the user asks about running Claude in a container, wants to isolate Claude's filesystem access, or needs a reproducible dev setup with Podman. Also use when the user mentions dev containers, Containerfiles, or sandboxed development workflows.

  • Fitness agent

    zxt6991-source/fitness-agent/skills/fitness-agent Skill

    no license0★ repo

    Operate the local Fitness Agent CLI for Strava and Obsidian workflows. Use when Codex needs to authenticate with Strava, list or fetch Strava activities, export workout data as Obsidian-compatible Markdown, configure a local Obsidian vault path, or write/sync workout notes into a vault from the local fitness-agent project.

  • Prompt leverage

    quanpersie2001/pulse/skills/prompt-leverage Skill

    no license0★ repo

    Use when the user asks to improve, structure, or templatize a prompt before execution by Codex or another AI agent.

  • Mcp builder

    JamesPrial/skills/skills/mcp-builder Skill

    no license0★ repo

    Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

    The same file is in 42 other repositories