AI & ML

yufanyang214-lab Bundle

Archive or recall durable facts, procedures, decisions, and events across Kimi Code CLI sessions using a standalone evidence/object/bright-dark-ledger workspace.

yufanyang214-lab Bundle

Archive an OpenClaw session into a standalone evidence/object/dual-ledger memory workspace, or recall prior sessions without using OpenClaw built-in memory.

marioluciofjr Bundle

Use SEMPRE que a pessoa usuária quiser um prompt-link — qualquer pedido para transformar um texto em link de prompt, gerar link de prompt, criar URL com o prompt preenchido, montar query string de um prompt, encurtar um prompt em link ou compartilhar um prompt como link, para Claude, ChatGPT, Perplexity ou Grok. Também quando mencionar prompt-link, claude.ai/new?q, chatgpt.com/?q, perplexity.ai/search/new?q, grok.com/?q, query string, percent-encoding, %0A, %20, codificar texto para URL ou encurtador zip1.io. Se faltar o prompt ou o destino, ou se o texto passar do limite, o fluxo avisa antes de montar qualquer link.

librefang Skill

Vector database expert for embeddings, similarity search, RAG patterns, and indexing strategies

librefang Skill

Machine learning engineer expert for PyTorch, scikit-learn, model evaluation, and MLOps

lcplyoohoo Bundle

Security audit workflow for vetting third-party Claude Skills, plugins, slash commands, agent definitions, hooks, and MCP configs for malicious content BEFORE they get installed or enabled. Use this whenever the user mentions installing, downloading, trying out, reviewing, or vetting any skill or plugin from a marketplace or repo (ClawHub, skills.sh, GitHub, npm, a pasted URL or zip), asks whether a skill is safe or trustworthy, says "check this skill out" or "should I add this" — even if they never say the words security or audit. Also use before enabling any skill file that did not originate on this machine, and when re-checking third-party skills that are already installed. Never install first and audit later — run this workflow first.

librefang Skill

LLM fine-tuning expert for LoRA, QLoRA, dataset preparation, and training optimization

librefang Skill

Prompt engineering expert for chain-of-thought, few-shot learning, evaluation, and LLM optimization

c1trusovo831 Bundle

Write or revise agent instructions, including skills, AGENTS.md, CLAUDE.md, and their supporting references.

librefang Skill

Use Xquik for X and Twitter social data workflows through its public API, SDKs, MCP server, webhooks, and installable agent skill.

librefang Bundle

Generate a structured meeting agenda from a topic and duration.

aryanyaya Bundle

Run protected AAS maintainer sweeps, PR merge batches, canonical sync, Core preview checks, and scripted releases. Use for repository maintenance, main alignment, CLI/MCP/Workbench changes, or release work; not ordinary contribution tasks.

mmacy Skill

Play a B/X dungeon crawl through the osrlib-referee MCP server, where the osrlib engine owns every roll and all game state and the LLM narrates and adjudicates freeform intent. Covers dungeon exploration, encounters, battle, the town economy, and level-up narration. Use when starting or continuing an osrlib-referee session, exploring a dungeon, running an encounter, resolving battle, or shopping and healing in town.

no1hoo Bundle

Transplants an existing prompt into a different domain, role, or subject while preserving its structural skeleton. Use when the user wants a "same-type" variant of a prompt — 同类提示词改版, 换领域, 换角色, 把 A 场景改成 B 场景, 移植提示词, adapt/rewrite a prompt for another industry, or asks to convert a scene/persona prompt (e.g. Japanese convenience store street corner → deep-sea aquaculture platform; shrimp nutrition → pufferfish nutrition; street food stall → space station canteen). Outputs a ready-to-paste prompt plus a change-mapping table and explicit judgement calls.

no1hoo Bundle

把模糊、不完整的提示词或需求，改写成清晰、具体、可执行的提示词。当用户说「帮我改进提示词」「优化提示词」「把需求说清楚」「我想让 AI 做某件事但不知道怎么说」「帮我写个提示词」「这个提示词行不行」，或提到「第一性原理」「亚里士多德模式」「proof-based」时使用。流程为先追问澄清缺口，再输出可直接复制的提示词与改动说明。Transforms vague prompts into clear, specific, actionable ones via clarifying questions before rewriting.

godimas101 Bundle

Use for Space Engineers asset creation: 3D models (Blender + SEUT addon, FBX → MwmBuilder → .mwm), texture channel packing (DX11 layout: _cm/_ng/_add/_alphamask), Havok collision meshes, material Technique values (MESH, DECAL, GLASS, HOLO, SHIELD…), LCD textures (PNG → texconv → .dds), and audio (WAV → xWMAEncode.exe → .xwm). Concrete triggers: making a custom block model, packing textures for a mod, setting up collisions, converting audio for the game. SKIP for: SBC-only content that reuses vanilla art (use se-sbc), C# code or scripts (use se-csharp / se-tss / se-pb-scripts), framework integrations without art work (use se-frameworks).

godimas101 Bundle

Use for Space Engineers server or client plugin development. Torch = dedicated server framework (plugins load into a headless server process, use TorchAPI, NexusV3 for multi-server). Pulsar = client-side plugin loader (plugins load into the game client via PluginHub, HarmonyLib for patching). Concrete triggers: 'write a Torch plugin', 'Pulsar plugin', HarmonyLib patch on a game method, TorchAPI usage, multi-server (NexusV3) coordination, dedicated-server-only modding without changing the game files. SKIP for: mods that ship with a saved game / world (use se-csharp / se-sbc / etc.), in-game PB scripts (use se-pb-scripts). Plugins are NOT mods — different install path, different lifecycle, different auth model.

mmacy Skill

Create B/X (Basic/Expert) characters and assemble a party through the deterministic, seeded chargen CLI, where the osrlib engine rolls every ability score, hit-point total, and gold piece — never the LLM. Use when creating a new character, rolling up a PC, or building a party before starting an adventure.

texarkanine Skill

Set up stockroom on this machine — check prerequisites, provision the per-machine torch build, smoke-test the embedding path, put the `stockroom` command on PATH, schedule the nightly ingest+embed, and run the first full ingest. Run this on a new machine, after a "run sr-initialize" error from a sibling skill, or any time the setup seems broken; it re-probes and only does what is still missing.

magomed-esendirov Bundle

Makes an AI agent itself perform the five daily Islamic prayers (salah/namaz: Fajr, Dhuhr, Asr, Maghrib, Isha) on location-based sun times from the free AlAdhan API, then re-arms a timer for the next prayer. Not a human reminder app — the agent prays. Works in any agent host that reads Agent Skills (SKILL.md). Use when a Muslim wants their agent to observe salah, asks to install or set up a namaz/salah skill, or wants to change the home city or madhhab of an agent that already prays.

jfear Bundle

Use when starting, planning, implementing, reviewing, or closing a non-trivial OpenSpec change that needs durable specifications and disciplined agent execution.

b7kompirine Bundle

GTA V / FiveM asset data and measured production rules — props, objects, doors, ytyp/ymap/MLO, LOD, grass, flags, particles (.ypt), shader/texture/light/timecycle/decal/parallax, clothing, vehicles; vanilla animation name and bone tag lookup. Use it to look at the real data BEFORE WRITING CODE in any prop, door, map, effect or look task, even if the user does not ask. Triggers — model name (prop_*, v_ilev_*, des_*), door does not open, object does not move, AddDoorToSystem, TaskPlayAnim, anim dict, bone tag, specialAttribute, flag number (1572872), LOD, disappears at distance, fxName, StartParticleFx, light does not turn on, TimeFlags, dark interior, timecycle, decal, parallax, DUI screen, fragment, RayFire destruction, freemode clothing, handling, modkit, Sollumz, CodeWalker. Use for GTA V / FiveM props, doors, maps, particles, materials, lights and clothing.

prompthon-io Bundle

Explain a business dataset with source-backed shape, grain, quality checks, distributions, numeric summaries, descriptive metrics and next questions. Use for read-only analysis; do not rewrite source tables, operate CRM entities or make unsupported causal claims.

prathameshlonare Bundle

Builds a resume from scratch by intake, or audits an existing resume/CV against four independent systems — traditional ATS keyword/parsing engines, AI-powered semantic search tools, human recruiter skimming, and AI-writing-pattern/credibility risk (generic-LLM phrasing that reads as templated to a human, independent of any detector) — against a target role or JD. Produces severity-coded findings, gap analysis, bullet rewrites, and an estimate-labeled match score for audits; a full drafted resume plus open placeholders for generation. Use when a user wants a resume/CV built or drafted from their information (no existing document), or uploads/describes an existing resume and asks for feedback, an ATS check, a "will this pass" review, a JD match, wants to know why they aren't getting callbacks, wants a recruiter's honest opinion, wants to compare resumes, or asks to make a resume sound "less AI" / "more human" / avoid sounding generated. Also trigger when a user uploads output from a third-party resume-scoring t

nyxandro Bundle

Property-based tests: fast-check with Vitest in TypeScript, Hypothesis with pytest in Python. Writes new property tests, exposes a property in code that seems to have none, reviews existing property tests for ones that assert nothing, and triages a failed property test into a wrong property, an ambiguous spec or a real bug. Use it when the owner asks for property tests or for inputs the existing tests miss: "write property tests for this parser", "find the edge cases this function misses", "test the invariants", "tests are green but prod breaks on odd input", "model-based test for this state machine or cache", "do these property tests check anything real", "my property test failed, is it a bug or a wrong property". Also use it on your own in test-first work when the function has an inverse, normalizes or sorts, does arithmetic over money or limits, parses data through a schema, or holds state. Not for binary fuzzing (libFuzzer, AFL), mutation testing, benchmarks or end-to-end UI tests.

prompthon-io Bundle

Plan and inspect prompt-cache behavior for long-running Claude agent loops. Use when a user wants to split stable tool, system, and history context into cacheable layers, compare captured cache metadata, estimate cost impact from supplied pricing inputs, or keep durable memory outside the cached prefix.

prompthon-io Bundle

Compare two structured agent-run artifacts to estimate cache efficiency, explain likely cache breaks, and produce a local benchmark report. Use when a user wants to understand whether a prompt layout, tool manifest, or history shape is helping or hurting prompt-cache reuse.

prompthon-io Skill

Review Gmail customer-support threads with a minimal-token, connector-first workflow. Use when Codex needs to read Gmail through the Codex Gmail connector, import cleaned messages into a local SQLite issue queue, apply deterministic cleanup and classification first, reserve model usage for JSON-only message understanding and draft-field generation, and support dashboard review, customer approval, and queued reply handling for a local single-operator support workflow.

prompthon-io Skill

Review inbound customer email, ground the reply in a local policy or FAQ document, and draft a safe response for complaints and queries.

mike22890 Bundle

MCP（Model Context Protocol）服务器的开发与集成。当需要创建 MCP 工具、开发 MCP 服务、对接外部 API（天气/数据库/搜索/支付）、让 AI agent 获得新能力、标准化工具接口、MCP TypeScript/Python SDK 使用、工具注册与发现时使用。

mike22890 Bundle

创建、测试、优化 agent skill。当需要开发新 skill、改进现有 skill、评测 skill 质量、编写 SKILL.md、设计 skill 架构、skill 版本管理、自动化测试、将重复工作流封装为 skill 时使用。也适用于 skill 发布到 GitHub 前的准备工作。

jetxu-llm Bundle

Complete non-trivial work once its objective, authorization, decision frame, and acceptance criteria are stable enough to act: choose the smallest fully working route, preserve required behavior, validate the actual outcome including the receiver boundary when material, and stop within scope. Use for multi-step, delegated, resumed, design, implementation, debugging, refactoring, migration, artifact, document, presentation, operational, or final-synthesis work whose main challenge is bounded execution and completion. If the central uncertainty is still the value function, problem boundary, causal or evidence model, decision owner, time horizon, or governing principle, stop convergence and return to framing before proceeding. Do not use for simple questions or trivial transformations.

prompthon-io Skill

审阅客户来信，以本地政策或 FAQ 文档为依据起草安全的投诉和咨询回复。

denizaslan36 Skill

Perform a final readiness audit of a repository, package, tool, model, dataset, or public artifact before release. Use when the user asks whether work is ready to publish, open-source, tag, package, or hand off. Do not use for ordinary feature completion or to publish anything without explicit authorization.

denizaslan36 Skill

Audit claims in a technical, scientific, analytical, or research document for evidentiary strength and calibrated wording. Use for report review, source-to-claim checking, uncertainty review, citation support, or distinguishing measurements from inference. Do not use for ordinary copyediting or to begin new research unless the user requests it.

hideki5123 Bundle

Run OpenAI Codex CLI from the terminal for batch / one-shot `codex exec` invocations with output captured to an `-o` file. Use for fire-and-forget runs, scripted CI, code review batches (`codex review`), and resuming a specific session by id (`codex exec resume`). Always use this skill instead of `mcp__codex__codex` and `mcp__codex__codex-reply` MCP tools — CLI is faster and more capable. For streaming UX, multi-turn dialogues that need live state, structured (JSON schema) output, attaching images, or programmatic event handling → **use codex-server instead**. codex-server uses the user's ChatGPT subscription via Codex App Server and decoupled async invocation that bypasses the Bash 2-minute timeout entirely. Trigger patterns (batch-only — chat/streaming triggers belong to codex-server): codex-cli / codex CLI / codex exec / codex review / codex resume / codex fork / batch + {codex, gpt} / one-shot + {codex, gpt} / non-interactive + codex / codex {to a file, -o, output file, captured output} / CI + {codex, gpt