agentsclimarketplace

Context & AI engineering

25,991 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

  • Agentskillshub cli

    zhuyansen/agentskillshub-cli Skill

    1★

    Use when the user wants to find, evaluate, audit, or install an open-source AI agent skill or MCP server — e.g. "find an MCP server for Postgres", "is this skill safe to install", "what should I use to scrape a website". Searches a quality-scored, security-graded catalog of ~20K skills locally (cached index, zero backend load) and returns each result's security grade, quality score, and install commands so you can check trust BEFORE installing.

  • Loop it

    fhajjej-ship-it/loop-it/skills/loop-it Skill

    1★ repo

    Choose, compile, launch, or run bounded AI-agent goals with explicit deliverables, proof rubrics, verifier gates, iteration caps, stop conditions, and approval boundaries. Includes a categorized Loop goals library for product, design, research, content, data, and operations plus advanced engineering loops for Codex, Claude Code, Cursor, and other SKILL.md-compatible agents. Use when the user says "loop it", asks for a loop goal, asks what should be looped next, or wants iterative improvement with proof. Do not use for simple one-shot edits unless the user asks for a repeatable loop.

  • Android project standard

    VoldemortGin/AI-Coding-Skill-Bible/android-project-standard Skill

    1★ repo

    Enforce a strict, model-agnostic, AI-navigable Kotlin/Android project standard: the Kotlin compiler's null-safety + sealed/`when` exhaustiveness as the static guarantee, plus `allWarningsAsErrors` + library-module `explicitApi()` + a detekt static-analysis gate as a one-command zero-warning gate, the escape hatches (`!!` non-null assertion / unchecked `as` / `lateinit` abuse / platform-type leaks / `GlobalScope`) shut via detekt-as-error, kotlinx.serialization + `@JvmInline value class` newtypes at boundaries (parse don't validate), a Gradle multi-module module-per-domain structure, a zero-SDK zero-framework `:domain` module behind provider interface seams with a default `MockProvider`, Hilt composition-root injection in `:app`, a version catalog pinning every version, an ktlint + detekt double lint gate, a CLAUDE.md in every module, and scaffold/conformance scripts. Use whenever starting or scaffolding a Kotlin or Android project; setting up Gradle / settings.gradle.kts / libs.versions.toml / detekt / ktlint / CI; deciding module structure; adding an LLM / embedding / vector-store dependency; wiring providers or adapters with Hilt; or checking that an existing Android project conforms. Apply it even when the user only says "start an Android project", "set up the structure", "add an LLM", or "wire up CI" without naming the standard. This skill governs the Kotlin host's internals; it is complementary to polyglot-core-standard, which governs the cross-language seam (generated UniFFI/JNI bindings).

  • Goal task prompt

    zitao4588-create/goal-task-prompt/skill/goal-task-prompt Skill

    1★ repo

    Use this skill when the user wants to turn a rough idea, project ask, debugging request, or large task into a concise, copy-ready Codex /goal prompt. Also use it for Loop Cards, delegation briefs, multi-agent orchestration, and prompt or skill evaluation when the user explicitly needs those advanced formats.

  • Memory card studio

    wuA-zs/memory-card-studio/memory-card-studio Skill

    1★ repo

    Use when the user asks to create or update persistent local memory-card projects, generate flashcards or spaced-repetition card libraries from files, URLs, web pages, or other readable sources, create a zero-install card-review frontend, start today's due-card review, or study an existing local card library in conversation.

  • Prompt engineering instructions

    bybaochaule/prompt-engineering-instructions Skill

    1★

    Use this skill to create, revise, evaluate, or standardize prompts and agent instructions using role, goal, context, examples, constraints, and reliability tests. Trigger when the user asks for prompt engineering, system or developer instructions, few-shot examples, prompt templates, agent behavior guidance, or prompt refinement. Do not use it for policy bypassing, hidden instruction extraction, unsafe automation, or unrelated copywriting.

  • Skillforge repo to skill

    xiaokillua/skillforge/skills/skillforge-repo-to-skill Skill

    1★ repo

    Use this skill when the task is to turn a GitHub repository or local codebase into a portable agent skill for Claude, Codex, GitHub Copilot, OpenClaw, Hermes, or other Agent Skills compatible runtimes.

  • Codex project memory

    Retourflow/project-memory-for-codex/skill/codex-project-memory Skill

    1★ repo

    Maintain a minimal, token-aware, verified project memory for Codex repositories. Use when initializing project context files, resuming work across sessions, changing architecture or project goals, recording decisions, updating current state after implementation, or reducing bloated and stale agent documentation.

  • Unified loop theory

    Agent-Pattern-Labs/unified-loop-theory/skills/unified-loop-theory Skill

    no license1★ repo

    Use when Codex is asked to create or run a master loop, unified loop, objective-driven control loop, recursive improvement loop, QA/eval loop, agentic QA system, release gate, synthetic test loop, production replay loop, trace capture plan, LLM judge stack, taste/product-quality review loop, or failure-diagnosis workflow. Helps Codex turn vague goals into objectives, worlds, probes, traces, judges, repairs, memory updates, gates, and the next loop iteration.

  • Npx skill install

    OpenGHz/npx-skill-install/npx-skill-install Skill

    1★ repo

    Install agent skills with `npx skills` the safe, no-duplicate-work way. Before installing, it checks (in one shot) whether the skill is already installed and whether every agent's skills directory is symlinked onto one shared hub — so you install ONCE instead of once per agent. Use this whenever the user wants to add, install, or set up a skill via `npx skills` / the skills CLI, hands over a skills repo or a skill name to install, asks why a skill isn't showing up in some agents, or wants to share one skills folder across Claude Code / Codex / Cursor / Gemini / Copilot / OpenCode. Triggers on: "install this skill", "add the X skill", "npx skills add", "set up this skills repo", "安装这个技能", "装一下这个 skill", "把这个技能装上", "用 npx skills 安装", "让所有 agent 都能用这个技能", "技能装了吗 / 是不是已经安装了", "把各个 agent 的 skills 目录软链接到一起".

  • Agentic build loop

    impactbrussels/AINativeOS/skills/agentic-build-loop Skill

    1★ repo

    Use when a founder says "how do I actually build this", "Claude Code/Codex/Cursor keeps making a mess", "the agent wrote spaghetti", "it passed but I don't trust it", "I'm just vibe-coding", "the agent went off and rewrote everything", "do I need tests for this", or is starting a feature and wants the agent on a leash. Triggers on plan mode, spec-first, acceptance tests, review the diff, agentic technical debt, iteration cap, runaway loop. Tool-agnostic across Claude Code, Codex, Cursor, Gemini CLI. Produces a spec plus acceptance tests, a plan-mode framework, then agent-filled logic gated on a green suite. Skill turns vibe-coding into a spec-and-test loop; without it, the agent builds the wrong thing fluently.

  • Gpt image 2 apikey skill

    junxxxxiao/gpt-image-2-apikey-skill Skill

    no license1★

    Generate or edit images through a separate API-key-backed GPT Image 2 provider without using Codex's own OpenAI/Codex credentials. Use when the user asks Codex to create, render, draw, generate, edit, modify, restyle, or transform an image with gpt-image-2/image-2 via an isolated API key stored in a private user config file, especially when the current Codex session is already authenticated with a different API key.

  • Agent spawner

    Razaib-khan/ForgeWeave/src/forgeweave/templates/claude/.claude/skills/agent-spawner Skill

    1★ repo

    Creates and configures subagents dynamically with defined roles, constraints, and tool permissions

  • Atlas agent

    wpfbcr/AtlasAgent/skills/atlas-agent Skill

    1★ repo

    Plans global travel itineraries with multi-plan comparison, day-by-day schedules, multi-currency budgets, visa/diet/transport guidance, and PDF export. Use when the user asks for trip planning, vacation itinerary, travel guide, 行程规划, 旅游攻略, or holiday planning for any destination worldwide.

  • Ax interface analysis

    Lightbridge-KS/agent-skills/plugins/coding/skills/ax-interface-analysis Skill

    1★ repo

    Analyze the agent experience ("AX") of an interface an AI agent drives — a CLI, MCP server, HTTP/REST API, or library/SDK — against the principles of what makes a system easy for a reasoning-loop agent to use, and prescribe prioritized fixes. Use this skill only when the user explicitly invokes it by name (`ax-interface-analysis`) or near-match mentioning.

  • Agent plan mode skill

    wildanfadh/agent-plan-mode-skill Skill

    1★

    Use this skill whenever the user asks for plan mode, planning mode, planning-first workflow, read-only analysis before implementation, or wants an agent/tool that lacks native plan mode to avoid making changes until a plan is approved. This skill is especially important when the user says phrases like "mode plan", "buat rencana dulu", "jangan edit dulu", "analisis dulu", "read-only dulu", "no changes yet", or asks to use an agent while preventing edits. It makes the agent inspect context safely, produce a clear execution plan, and wait for explicit user approval before modifying files, running write operations, committing, pushing, or changing system state.

  • Agent teams

    HorizonBrute/Standardized_AI_Looping_Language-SAILL/skills/agent-teams Skill

    1★ repo

    Create or edit agent-team definitions in local.agent_teams.md at any scope (OS, project, brain, or subfolder) — define the agent chain, each role's model-group preference, and loop/retry constructs. Use when the user types /agent-teams, asks to "add an agent team", "define an agent team", "edit agent teams", or "create a team for this project/scope".

  • Skill forge

    truthandstories07-sudo/skill-forge Skill

    1★

    Create, evaluate, and improve agent skills (SKILL.md format) using an evidence-based methodology distilled from 244 top skills across the Anthropic, OpenAI, vendor, and community ecosystems. Use when the user wants to turn a repeated task, workflow, or expertise into a skill, asks to create or scaffold a SKILL.md, wants an existing skill judged, scored, or audited, or asks to improve a skill's structure, description, or triggering (中文触发:做成技能、创建 skill、沉淀为 skill、写一个技能、评估/评判这个 skill、优化 skill 描述、skill 体检). For routing between already-installed skills use agent-skill-hub instead; for Lark API skills, lark-skill-maker supplies the domain template while this skill governs quality and testing. NOT for writing ordinary prompts, CLAUDE.md/AGENTS.md rule files, or subagent definitions.

  • Skillopt skill writer

    Waylish/skillopt-skill-writer Skill

    1★

    Create, improve, and validate agent skills using a SkillOpt-inspired workflow: scored evidence, bounded add/delete/replace edits, validation gates, rejected-edit memory, and compact deployable SKILL.md artifacts. Trigger when the user says "write a skill", "create a Codex skill", "make a Claude skill", "turn this workflow into a skill", "optimize this skill", "improve SKILL.md", "skill quality", "best_skill.md", "use SkillOpt", "validation-gated skill", "write a skill for writing skills", "写skill", "创建Codex技能", "优化技能", "技能文档", "写一个 skill", "把流程变成skill", or provides a repeatable workflow that should become a reusable agent procedure.

  • Claude cabinet

    gengshirong1128-boop/claude-cabinet Skill

    1★

    Cost-aware Claude model routing across Fable 5, Opus, and Sonnet. Use when a task needs Claude Code model selection, premium-model cost control, model delegation, explicit escalation gates, Fable/Opus review, Sonnet execution, or a reproducible routing and cost report.

  • Claude from codex

    yujingz/use-claude-code-from-codex/skills/claude-from-codex Skill

    no license1★ repo

    Delegate explicit Codex tasks to the local Claude Code CLI for parallel investigation, review, second opinions, and bounded implementation slices.

  • Init project

    tuannv0069/claude-doc-genome/skills/init-project Skill

    1★ repo

    Use this skill ONCE per new project to deploy the agent documentation standard — the trio CLAUDE.md + .claude/rules/ + docs/agent-guide/ — plus the skill-authoring set (skill-designer, skill-writer, skill-writer-auditor) and the document-writer documentation skill. Trigger phrases — "init project docs", "khởi tạo bộ tài liệu agent", "deploy doc standard", "/init-project". Two maintenance modes run ONLY at this master repo — "/init-project check" (drift report) and "/init-project promote" (consolidate live → bundle). Do NOT use to author a single new skill (use skill-designer/skill-writer), and do NOT re-run init on an already-initialized project.

  • Documentation workspace skill

    SBTopZZZ-LG/documentation-workspace-skill Skill

    no license1★

    Use this skill whenever the user asks to bootstrap, initialize, set up, scaffold, or create a documentation workspace, docs folder, knowledge base, or note repository in an empty or near-empty directory. Triggers on phrases like "set up a docs workspace", "create documentation structure", "initialize a new docs project", "organize my notes into folders", "I want a new knowledge base", "scaffold a docs folder", or any variant the user uses to express wanting a fresh, organized, file-based documentation system. This skill creates a complete, versioned, self-contained documentation workspace with AGENTS.md, README.md, index.yaml, .gitignore, package.json, and helper scripts (format, validate, hash, reindex). It does NOT cover editing an existing workspace (the workspace's own AGENTS.md governs that), GitHub repository creation or hosting, or rendering the workspace as a UI. If the target directory already contains an AGENTS.md, the workspace is already bootstrapped — do not re-run this skill, just read AGENTS.md and follow it.

  • Handoff tracker

    knowledgeco/handoff-tracker Skill

    1★

    Structured decision handoffs for multi-model project orchestration. When a model finishes a turn or hands work off — especially when the next step is unclear to the operator, or the operator cannot read the code themselves — the outgoing model must not leave a dead end. It must surface the viable next-step OPTIONS it can see, make an explicit RECOMMENDATION with honest tradeoffs, and leave the CHOICE to the operator. Each handoff records the option set, the recommendation, the operator's choice, and the eventual outcome — which simultaneously keeps the operator oriented when the next step is unclear, becomes a recovery map when a chosen path fails (the untaken options are still sitting at the fork), and forms a provenance trail of exactly where human judgment steered. Use this whenever handing off work between models or sessions, whenever the next step is unclear or consequential, whenever an operator who can't read the code needs the fork framed as a decision they can actually make, or whenever a model is about to stop and owes the operator a "here's what's next" — even if the user doesn't say "handoff," "options," or "decision."

  • Multi harness

    LFTPadilla/agent-dev-kit/overnight-task-kit/skills/multi-harness Skill

    1★ repo

    Delegate bounded coding-agent work through another local harness such as Pi or OpenCode, or through the local Codex CLI when the user explicitly requests a Codex-native model that the current session's subagent API does not expose. Use for explicit cross-harness requests, harness comparisons, external-only runtimes, or named Codex models such as GPT-5.3 Codex Spark that require a CLI fallback. Do not use for generic requests to spawn, delegate, fan out, orchestrate, or parallelize with models already exposed by the current harness.

  • Paniolo config init

    paniolo-ai/scan/.agents/skills/paniolo-config-init Skill

    1★ repo

    Scaffold a starter paniolo.config.json by asking the user which AI vendors/harnesses (Claude Code, Cursor, Copilot, Codex, Gemini, Antigravity) this repo should support. Use when asked to init, scaffold, or set up paniolo.config.json, configure which AI vendors a repo supports, or on /paniolo-config-init.

  • Coding agent handoff

    Aaryan-Kapoor/coding-agent-handoff Skill

    1★

    Turn the current chat conversation into a self-contained markdown brief (a saved file or artifact) that a coding agent — Claude Code, Codex, Cursor, or similar — can execute in the user's repository. Use when the user asks to hand this off to a coding agent / Claude Code / Codex, write it up as a task or spec to build, or otherwise move the work from this chat into a coding tool to implement.

  • Agent codebase analysis skill

    yuecao365/agent-codebase-analysis-skill Skill

    1★

    Use when analyzing an LLM-agent repository, agent harness, single-agent runtime, multi-agent system, agent framework, or mature open-source agent project to understand or present its system design, runtime behavior, context/state handling, tools, permissions, memory, RAG, events, artifacts, observability, evals, coordination model, Mermaid architecture diagrams, or plain-language HTML architecture report.

  • Yapui

    Tatendaz/yapui Skill

    1★

    View, preview, open, render, or serve any HTML file/page/mockup/prototype in the browser with a live two-way feedback loop. Use this whenever the user wants to look at HTML in a browser, see a rendered page, preview a mockup or prototype, or iterate on an HTML UI — instead of a bare `open file.html` (file:// blocks the mic + screen capture and gives no feedback channel). It serves the HTML from a local relay, injects a feedback widget (typed notes, voice dictation, screen recording, screenshots, and click-to-pick element selection), and boots a resident pre-warmed Claude agent that picks up each note the instant it is sent (~0s pickup), applies the fix, streams live "what I'm doing" status to the page, and replies in the browser. Trigger for ANY request to view / preview / serve / open / render HTML.

  • Agentctx

    rahulkolekardev/agentctx/skills/agentctx Skill

    1★ repo

    Make a repository ready for AI coding agents by scanning project facts and generating AGENTS.md, CLAUDE.md, GEMINI.md, Copilot instructions, Cursor rules, repo maps, doctor reports, and task-specific context packs. Use when preparing repos for Claude Code, OpenAI Codex, Gemini CLI, Google Antigravity, GitHub Copilot, Cursor, Windsurf, Amp, Aider, Cline, Roo Code, Continue, OpenHands, Replit Agent, Vercel Skills, or other coding agents.

  • Agent bus delegate

    adjcjh777/agent-bus-skills/skills/agent-bus-delegate Skill

    no license1★ repo

    Send Agent Bus messages, delegate tasks, verify visible delivery, and reply or ACK to Bus messages. Use when the user asks to send, delegate, dispatch, notify, wake, acknowledge, or reply to another agent or thread; handle pending targets; use reply_message or agent-bus reply; or distinguish durable Bus storage from visible Codex thread delivery.

  • Who am i skill

    Pointa-Labs/who-am-i-skill Skill

    1★

    Analyze the user's OWN messages across all their local AI conversations (Claude Code, Codex, VS Code Copilot, Cursor, and more) to produce a sharp, evidence-grounded "Who Am I" self-portrait — personality, strengths, blind spots, and concrete ways to grow. Use this WHENEVER the user wants to "know themselves" / 认识自己 through their AI chat history, asks for a personality read / 性格分析 / 自我画像 / 自我反思 based on how they talk to AI, wants a report on "what kind of person am I" drawn from past conversations, or says things like "analyze my chat history and tell me who I am", "我想通过我和AI的 对话来了解我自己", "看看我是个什么样的人", or "who am I based on my prompts". Reads only local data; nothing is uploaded.

  • Coalboard

    TheColliery/CoalBoard/plugin/skills/coalboard Skill

    1★ repo

    Consensus & debate board. AUTO-trigger = the error-not-allowed slice (security/crypto, DB/financial migrations, high-precision math, anything catastrophic-on-error); convened MANUALLY ("/coalboard") it generalizes to any hard problem in ANY domain worth several diverse perspectives — code, docs, math, research, translation, legal. With the user's consent it convenes diverse epistemic lenses (empirical/source-grounded, formal/logical, a show-me skeptic) debating IN PARALLEL; a judge synthesizes on VERIFIED inputs; an independent out-of-frame solver breaks ties; the human signs off. Guarantees bounded cost (no whack-a-mole) and zero-breakage (staging) — it improves correctness, it does not claim a reliability number. Off ~90% of the time. Triggers: "/coalboard", "convene the board", a critical-task signal, or a CoalTipple escalation hand-off. Cross-agent (verified: Claude Code + Antigravity; others designed-for, unverified; Claude Code adds cost-optimized tiering). Zero-dependency, offline, no API keys.

  • Local memory skill

    yaoxiao99/local-memory-skill Skill

    1★

    Use when an agent needs to set up, read, write, or maintain a fully local Markdown-based memory system; when the user says "remember this", "save to memory", "set up local memory", or wants cross-session context without a cloud memory provider.

  • Auto polish skill

    Shavy72/auto-polish-skill Skill

    1★

    Mandatory final phase for every new UI/frontend build (feature, page, component, redesign, full app) — runs automatically, without the user asking. Mini-addons (<~30 changed lines) only on demand via /auto-polish. Runs a dynamic multi-agent workflow with fixed amplification stages 5x → 10x → +3x and design judges in between, including a resource inventory (which installed skills/MCPs are still unused?), sign-off from every angle (desktop, mobile, code, video for animations) — local first, then live after shipping. Triggers (DE+EN): feature finished, polish phase, 'auto-polish', 'polish das', 'maximal geiler machen', 'showcase level', 'design steigern'.

  • Ha automations

    t3chnaztea/home-assistant-skills/skills/ha-automations Skill

    1★ repo

    Use when writing, editing, debugging, or verifying Home Assistant automations with an agent: "write an automation that...", turn an English description into automation YAML, edit automations.yaml over SSH, an automation that never fires or fired once and went dead, a service call that silently did nothing, testing with automation.trigger, or reloading automations without restarting Core. Covers the write → check → reload → verify loop and the silent-failure traps (event-entity triggers, dead notify services, no-op writes). Not for the initial connection (ha-connect) or price/LLM-driven triggers (ha-external-triggers).

  • Skill manager

    potocibi/skills-manager/skills/skill-manager Skill

    1★ repo

    Manage, audit, deduplicate, merge, route, and APPLY the user's installed Claude Code skills, agents, commands, and plugins. Use when the user says "audit my skills", "dedupe skills", "merge skills", "skill registry", "which skill handles X", "skill health", "skill intake", "propose a skill", "should this be a skill?", "turn this into a skill", or BEFORE creating any new skill or agent (gap-check first). Also use when the user asks in plain English to run/use/apply whatever skill or agent handles a task ("use the right skill for this", "run the reviewer on this file", "apply whatever handles X") — resolve it from the registry and invoke it, chaining multi-step asks. Also use when the user shares a GitHub repo or folder of skills to compare against or import from ("is this skill better than mine", "check this skills repo", "add skills from this repo"), and to restore archived skills or fix registry drift.

  • Codex internal agent orchestration

    FairladyZ625/codex-internal-agent-orchestration Skill

    1★

    Coordinate complex coding and knowledge-work tasks inside a user's target repository with Codex's current main agent and in-task internal subagents. Use for multi-agent task graphs, parallel exploration, disjoint implementation slices, long-running supervision, evidence collection, adversarial review, final integration, and Harness Anything task/decision/fact workflows when the target repository has adopted Harness Anything.

  • Claude multi agent

    SYMBaiEX/skills/skills/claude-multi-agent Skill

    1★ repo

    Hand off engineering work to Claude Code as an autonomous team or native dynamic workflow, with explicit Opus/Sonnet routing, foreground completion barriers, resumable sessions, repeatable JavaScript workflow phases, hooks, and worktree isolation. Use when another agent or CI needs Claude Code to implement a scoped task, run a codebase-wide audit or migration, execute research-build-verify-gap-close phases, launch work in the background, or install project subagents and `.claude/workflows`.

  • Procraft

    Xerneas216/ProCraft/skills/procraft Skill

    1★ repo

    Use when the requested deliverable is an AI or model Prompt, prompt text, or 提示词, including an LLM, Agent, image, video, or audio Prompt, or a system/developer/user message, tool policy, Structured Output contract, or reusable AI workflow specification; do not use for direct execution of writing, research, coding, summarization, email, image generation, or any other underlying task unless the user asks for the Prompt or instructions that will perform it.

  • Harness engineering init

    Kinghonga/harness-engineering-init Skill

    1★

    Initialize and operate a project's AI-agent harness-engineering loop so it actually runs, not just gets scaffolded. Scaffolds the .harness/ structure, checks whether the loop is alive, drives task checkpoints, failure-trace recording, and the Critic analysis that converts traces into failure-pattern memory and rule updates. Use when the user says "harness", asks to set up / check / run harness engineering, mentions .harness/ or AGENTS.md discipline, wants an AI coding agent's loop to self-improve, or when checkpoints/traces/memory are not being maintained. Makes silent harness failure visible and closes the trace-to-Critic-to-failures loop that usually dies.

  • Agentic dev loop

    dimsedra/moatcraft-skills/.agents/skills/agentic-dev-loop Skill

    1★ repo

    Orchestrate the continuous agentic development loop. Routes between brand-product-alignment, front-end-designer, backend-architect, progress-mapper, implementation-tdd, alignment-audit, code-review, and explain-and-teach. Manages the state machine, cascade protocols, rollback, and Loop Summary. Use when guiding a feature, refactor, or iteration through the full development loop. Do NOT use as a direct implementation skill (delegate to implementation-tdd instead).

  • Agent stack

    14BryanEspinoza/agent-stack Skill

    1★

    Meta-skill de orquestación — cómo el agente debe pensar, detectar, decidir y actuar en proyectos frontend, evitando redundancia y alucinación.

  • Prompt triwizard

    HiMyNameIsDavidKim/prompt-triwizard-skill/skills/prompt-triwizard Skill

    1★ repo

    Use when writing, refactoring, or evaluating LLM prompts (system prompts, user prompts, few-shot examples, or tool-call schemas) for agent pipeline nodes or stages. Triggers on: "write a system prompt for this agent", "improve this LLM instruction", "evaluate this prompt", "add few-shot examples to this node". Does NOT trigger for general coding, debugging, tests, or non-LLM tasks.

  • Add skill

    Paldom/github-skills/.claude/skills/add-skill Skill

    1★ repo

    Authors or improves an Agent Skill in this repo via the eval-first workflow - scope one purpose, write trigger evals, draft SKILL.md, validate, iterate. Use when the user asks to create, add, write, refine, or fix a skill ("add a skill for X", "improve triggering of Y"). Not for scaffolding a new repository or editing infrastructure (hooks, CI, the validator/scripts, docs).

    The same file is in 11 other repositories

  • Agent config reference

    2702207741-dev/agent-skills-pipeline/agent-config-reference Skill

    1★ repo

    Use when the user asks about config.yaml, CLAUDE.md, .cursor/rules, .codex config, agent setup, or skill installation. Use when the user says 配置agent, 安装skill, 迁移skill, or needs to understand platform-specific config differences across Hermes Agent, Claude Code, Codex CLI, and Cursor.

  • Bootstrap sdd tdd

    kxdds/sdd-tdd/skills/bootstrap-sdd-tdd Skill

    1★ repo

    Use when a project needs the OpenSpec + Superpowers sdd-tdd (spec-to-plan-to-TDD) workflow checked, installed, and configured -- or cleanly removed -- in any coding agent (Codex, Claude Code, Cursor, Antigravity, OpenCode, Gemini CLI, Copilot CLI, or others).

  • Agent llm

    ksed8/cc-loopkit/.claude/skills/agent-llm Skill

    1★ repo

    Patterns for building on the Claude/Anthropic API — agent loops, tool design, prompt structure, structured output, context management, prompt caching, and evals. Use when writing code that calls Claude, designing an agentic loop, defining tools for an LLM, building RAG or an LLM-judge, or debugging refusals, truncation, streaming, or tool-call failures.

  • Deai zh gpt style

    YanhanLi/deai-zh-gpt-style/skills/deai-zh-gpt-style Skill

    1★ repo

    Use when editing Chinese or mixed Chinese-English/number text to reduce GPT-5.5/GPT-5.4 style habits, formulaic AI phrasing, over-structured wording, unnatural chat phrasing, or unwanted spaces between Chinese and English/digits.

  • Skillxray

    munzzyy/skillxray/skills/skillxray Skill

    1★ repo

    Scan an AI agent skill, plugin, or MCP bundle for security and hygiene problems before installing or trusting it - prompt injection in the instructions, hidden/invisible Unicode, dangerous commands (curl-pipe-sh, reverse shells), data exfiltration, hardcoded secrets, and auto-running hooks. Use before adding a third-party skill or plugin, before wiring up an MCP server someone else wrote, or whenever you're about to trust a SKILL.md you didn't write. Returns an A-F grade and a per-finding report.

  • Clics

    P-Essonam/clicsdev/skills/clics Skill

    1★ repo

    Query Clics analytics and manage projects, goals, funnels, and sessions through the Clics MCP server (`@clicsdev/mcp`), with the `clics` CLI (`@clicsdev/cli`) as a fallback. Use when a user asks about their analytics, mentions Clics, works with the Clics toolset, verifies Clics manually, or needs to choose between MCP and CLI.

  • Writing gpt 5 6 prompts

    CodingCossack/writing-gpt-5-6-prompts Skill

    1★

    Use when creating, rewriting, optimising, reviewing, or handing off prompts, system prompts, developer instructions, agent briefs, delegation prompts, task specifications, or reusable prompt templates for GPT-5.5 or GPT-5.6, especially when wording is procedural, repetitive, over-scaffolded, vague about success, or unclear about evidence, autonomy, tools, validation, output, or stopping.

  • Refine prompt

    basuev/pi-prompt-refiner/skills/refine-prompt Skill

    1★ repo

    Refines a rough prompt into a precise, ready-to-send Codex prompt. Use when the user asks to refine, improve, rewrite, optimize, or make a prompt more effective.

  • Opencode commander

    stone623101-cloud/opencode-commander/plugins/opencode-commander/skills/opencode-commander Skill

    1★ repo

    Orchestrate multiple OpenCode CLI workers from Codex, Claude Code, Cursor, or another coding agent. Use when the user explicitly asks to split a complex research, analysis, planning, review, or implementation task across OpenCode agents; requests parallel OpenCode execution; wants iTerm split-pane visibility; or asks the host agent to act as the overall coordinator and synthesize worker results.

  • Agent lean safe coding

    AgentPilotLab/agent-lean-safe-coding Skill

    no license1★

    Windows-safe lean coding workflow for Codex, Claude Code, and other AI coding agents. Use for coding, refactoring, bug fixing, dependency choices, code review, or implementation planning when the task may involve Windows shells, PowerShell, cmd, Git Bash, WSL, Unicode or Chinese text, Markdown or prompt files, UTF-8/GBK/CP936 encoding, CRLF/LF line endings, path portability, dependency reuse, YAGNI decisions, over-engineering risk, generated folders, small patches, or minimal verification. Also use when the user asks for lean code, safe Windows edits, smallest correct implementation, dependency review, anti-bloat review, or diff safety.

  • Ai incident analyst

    anmolg1997/prepostmortem-skills/ai-incident-analyst Skill

    1★ repo

    Use when analyzing failures of AI/LLM/agentic systems, an agent took a destructive or unintended action, a chatbot hallucinated with real-world consequences, prompt injection or jailbreak, model-quality or behavior regression after an update, reviewing an agent system design for failure modes before launch, writing a postmortem for an AI incident, or when the user mentions AI incidents, agent safety review, guardrails, prompt injection precedent, or "has an AI system done this before".

  • Agentify

    avnath13/agentify/agentify Skill

    1★ repo

    Design production-grade agentic AI systems from a natural-language use case. Runs a clarification loop, walks grounded decision trees (workflow vs agent, RAG vs fine-tuning, single vs multi-agent, autonomy tiers), and emits a detailed enterprise system design document with embedded interactive architecture, sequence, and data-flow diagrams. Every decision cites authoritative sources (Anthropic, OpenAI, OWASP, NIST, cloud well-architected frameworks, peer-reviewed surveys). Supports a production mode (buildable design) and an interview mode (system design interview framing with tradeoffs and probes). Use when the user wants to design, architect, review, or plan an AI agent, agentic workflow, RAG system, LLM application, copilot, or GenAI product, or asks "how should I build" an AI-powered system, or wants agentic system design interview practice.

  • Fde copilot

    inbusiness23/fde-skill/skills/fde-copilot Skill

    1★ repo

    Start here for a practical forward-deployed engineering workflow. Help a small or midsize organization improve one costly workflow at a time, coordinating discovery, solution design, safe deployment, and value review with a cautious 30-day payback target. Use for FDE work, workflow improvement, AI-agent project selection, rollout planning, ROI review, or deciding whether a process needs AI at all.

  • Project orchestrator

    code-wahaha/project-orchestrator-skill/skill/project-orchestrator Skill

    1★ repo

    Guide any software project from an idea or existing codebase to an approved specification, governed implementation, verification, and handoff. Use for creating, extending, migrating, refactoring, or taking over mini programs, web apps, mobile or desktop apps, APIs, backend services, automations, and other software; for both 0-to-1 projects and 1-to-100 existing or legacy projects. Trigger when a user says they have an idea, wants to build or change a product, asks to plan before coding, needs project rules or architecture, wants an existing repository audited, or wants guided, assisted, or automatic project execution with reusable preferences.

  • Resume handoff

    sduhantao-maker/resume-handoff Skill

    1★

    Use when work won't finish before the quota or context runs out and must be paused and handed off — triggers include 额度快用完, 额度用完前干不完, 续命, 交接, 生成交接文档, handoff, context running low, 上下文快满, pausing mid-task, or another agent must pick up and continue unfinished work and recover context.