AI & ML

getgrille Skill

Use when performing git operations on a repository via Grille. WHEN: 'git status', 'what changed', 'commit history', 'git log', 'show diff', 'what files changed', 'commit these changes', 'switch branch', 'create branch', 'stash changes', 'fetch from remote', 'where does this repo push to', 'what is the origin URL', 'git remotes', 'who wrote this line', 'who last changed this', 'git blame', 'per-line authorship', 'which commit changed this', 'git_status', 'git_log', 'git_diff', 'git_branches', 'git_show', 'git_stash_list', 'git_commit', 'git_checkout', 'git_stash', 'git_stash_pop', 'git_fetch', 'git_remote', 'git_blame'. DO NOT USE WHEN: 'push to remote' (git push not supported — SSH agent OS limitation; use process_run with git.exe manually); 'clone a repository' (out of scope); 'merge or rebase' (interactive, not supported).

luanquangminh Bundle

Single-agent research methodology — the search-fetch loop, query templates, GitHub patterns, and retrieval tools. Loaded by researcher agents spawned from lead-researcher. Do NOT use this skill directly for research — use lead-researcher instead, which orchestrates agents with configurable depth (low/medium/high/max). This skill is methodology, not orchestration.

luanquangminh Bundle

Unified heavy thinking orchestrator. Classifies problem type → dispatches to the right parallel thinking pattern. Modes: brainstorm (divergent ideation), solve (verifiable answers), decompose (break complex problems), unstick (reframe blockages). All modes use parallel agents with model=opus for maximum reasoning depth, then YOU synthesize. Actions: think deeply, brainstorm, decompose, solve, unstick, analyze. Keywords: heavy think, heavy thinking, think harder, brainstorm, decompose, break down, stuck, unstick, solve, parallel thinking, multi-perspective, think about this, explore deeply, reason about, analyze deeply, heavyskill, heavy-brainstorm. Use when: user says 'heavy think', 'think harder', 'brainstorm deeply', or any problem deserving more than shallow first-pass thinking — strategic decisions, hard technical problems, complex decomposition, creative exploration, feeling stuck. Do NOT use for: simple lookups, mechanical tasks, clear next steps, tasks needing implementation not thinking.

tt-james Bundle

Build, inspect, and reuse project knowledge graphs with Understand-Anything before coding. Use when entering a new repository, onboarding Codex or subagents to an unfamiliar codebase, creating a visual or chat-based project map, reducing repeated repository scans, planning multi-agent work, reviewing change impact, or needing a durable `.understand-anything/knowledge-graph.json` context.

zeroix07 Bundle

Implement large language model (LLM) chat completions using the z-ai-web-dev-sdk. Use this skill when the user needs to build conversational AI applications, chatbots, AI assistants, or any text generation features. Supports multi-turn conversations, system prompts, and context management.

getgrille Bundle

Use when evaluating Grille's security model, understanding permission boundaries, or assessing Grille for enterprise deployment. WHEN: 'how does Grille handle security', 'what can Claude not do', 'security model', 'prompt injection', 'audit guarantees', 'enterprise evaluation'. DO NOT USE WHEN: you need to use a specific tool — load the relevant module skill instead.

luanquangminh Bundle

MANDATORY entry point for ALL research tasks. Orchestrates researcher agents with configurable depth modes (low/medium/high/max). Determines languages, iterations, agent count, and sub-topics — then spawns parallel researcher agents and synthesizes results. Triggers: any research request, 'research X', 'what's the best', 'compare X vs Y', 'how should I approach', 'evaluate X', 'deep research', 'comprehensive analysis'. ALL research goes through lead-researcher first — never spawn researcher agents directly. Arguments: [mode] [topic] — e.g. 'high compare auth solutions' or 'max full ecosystem survey of MCP servers'.

miaoge-ge Skill

Expert Retrieval-Augmented Generation: chunking, embeddings, vector/hybrid search, reranking, and grounded answers. Trigger keywords: RAG, retrieval, embeddings, vector database, chunking, reranking, hybrid search, BM25, grounding, citations, hallucination, context window, recall. Use to build or debug RAG pipelines and improve answer quality/faithfulness.

zeroix07 Bundle

Generate podcast episodes from user-provided content or by searching the web for specified topics. If user uploads a text file/article, creates a dual-host dialogue podcast (or single-host upon request). If no content is provided, searches the web for information about the user-specified topic and generates a podcast. Duration scales with content size (3-20 minutes, ~240 chars/min). Uses z-ai-web-dev-sdk for LLM script generation and TTS audio synthesis. Outputs both a podcast script (Markdown) and a complete audio file (WAV).

miaoge-ge Skill

Expert TypeScript: type system, generics, narrowing, inference, and strict-mode safety. Trigger keywords: TypeScript, types, generics, type narrowing, tsconfig, strict, discriminated union, conditional/mapped types, utility types, declaration files, type error, satisfies, infer, any vs unknown. Use when writing/refactoring TS, fixing cryptic type errors, designing type-safe/generic APIs, or configuring the compiler.

miaoge-ge Skill

Use this skill when user needs to design LLM evaluation strategies, build test datasets, or ensure model quality. Trigger keywords: LLM testing, model evaluation, prompt engineering, regression testing, red teaming, hallucination detection, RAG testing, Agent testing, A/B testing, LLM-as-a-Judge. Applicable to quality assurance in model development and application deployment.

dsdsddd1 Bundle

Use when Codex needs authoritative Blender help, especially Blender Manual/API questions, Geometry Nodes behavior, node socket names, Blender MCP server usage, scene inspection, modifier/node-tree analysis, or implementation guidance that must match the active Blender version. Prefer official Blender documentation and live Blender introspection over memory; use this for building, debugging, or documenting Blender Python, Geometry Nodes, materials, modifiers, and add-ons.

bin1874 Bundle

创建高质量 MCP（模型上下文协议）服务器的指南，使 LLM 能够通过精心设计的工具与外部服务交互。在构建 MCP 服务器以集成外部 API 或服务时使用，无论是 Python（FastMCP）还是 Node/TypeScript（MCP SDK）。

miaoge-ge Skill

Use this skill when user needs to design deep learning models, perform training optimization, interpret papers, or solve engineering practice issues. Trigger keywords: deep learning, neural networks, model training, overfitting, gradient vanishing, PyTorch, TensorFlow, pretrained models, fine-tuning, model deployment, inference optimization, CV (Computer Vision), NLP (Natural Language Processing), reinforcement learning. Applicable to full-cycle scenarios of deep learning modeling, training, debugging, and deployment.

miaoge-ge Skill

Expert prompt engineering: prompt structure, few-shot, chain-of-thought, structured output, and iteration. Trigger keywords: prompt, prompt engineering, system prompt, few-shot, chain-of-thought, output format, JSON schema, structured output, instructions, role, examples, eval. Use for writing, debugging, or optimizing prompts for LLMs.

tt-james Bundle

Manage Codex and agent skill libraries by combining the best ideas from open skill ecosystems. Use when the user asks to audit installed skills, discover better skills on GitHub, choose active vs backup skills, deduplicate broad workflow skills, compare skill managers such as openai/skills, vercel-labs/skills, ComposioHQ/awesome-codex-skills, yeasy/ask, Agent-Skills, oh-my-skills, xiaobai-skills, install or update skill sets, create a skills inventory, publish skills to a GitHub skills repository, or design a team-ready skill governance workflow.

tt-james Bundle

Orchestrate CodeGraph and Understand-Anything for project knowledge graphs before coding. Use when entering a new repository, onboarding Codex or subagents, choosing between semantic code search and visual/chat project maps, reducing repeated repository scans, preparing multi-agent context, analyzing impact, or reusing durable project graph summaries.

pyro17ai Bundle

Practical guide for creating, editing, animating, compositing, keying, masking, tracking, applying effects, and exporting Adobe After Effects projects through an After Effects MCP bridge. Use when Codex needs to operate Adobe After Effects via MCP tools, plan AE compositions, automate layers/keyframes/effects, troubleshoot the after-effects-mcp bridge, or translate video-editing/VFX requests into concrete After Effects actions.

euroconic Skill

Take an AI feature spec and produce a full eval harness - extracted behaviors, golden + edge + adversarial test cases, a judge rubric, and a scorecard. Use before shipping any LLM-powered feature, when "it works on my examples" isn't enough, or when stakeholders ask "how do you know it's good". Triggers on phrases like "build evals for this feature", "score this AI feature", "test cases for LLM feature", "eval rubric", "AI quality scorecard", "оцени AI-фичу", "сделай эвалы".

sonia420116xh-max Bundle

Generate Sonia's Chinese weekly engineering report from configured local git repositories. Use when the user asks to generate, rewrite, schedule, automate, or explain the weekly report workflow, especially for AI MAX / Paid Collab work, this-week.prompt.txt, weekly-report-final-YYYY-MM-DD.md, or Friday weekly reporting.

coopersimson96 Bundle

Run hedge-fund-grade technical analysis on any stock or crypto ticker using the Kronos open source foundation model (trained on 12B candlesticks from 45 exchanges). Pass a ticker, get back a price prediction with confidence interval in plain English. Triggers on "/kronos", "run kronos on X", "kronos analysis", "predict X with kronos", "technical analysis on X", "what does kronos say about X".

jawwadfirdousi Skill

Summarizes the current git repository — branch, status, recent commits, and file mix. Use when the user asks for an overview of the repo, what kind of project this is, what's been happening recently, or to brief the agent on a new codebase.

davideldersveld Skill

Captures a complete telemetry log of a Cowork session — request interpretation, routing decisions, task graph, tool-call ledger, subagent dispatches, execution metrics, detected issues, cross-output consistency, latency/token outliers, and known unknowns — into a single markdown deliverable at output/session-diagnostic.md. Designed to be model-agnostic, topic-agnostic, and input/output-agnostic so the same skill applies to any future session. Use when user asks to "log diagnostics", "diagnostic log", "diagnostic report", "track this session", "trace the session", "session telemetry", "capture execution stats", "execution trace", "audit this run", "what did the agents do", "what tools did you use", "what happened in this session", "build a running log", "tell me everything you do", or pairs another request with "and also track everything" / "and create a diagnostic file" / "and log all the activity". Do NOT use for: TaskCreate-style live progress tracking (the task system handles the spinner UX), conversation

tianyiwang630-ship-it Bundle

Python-based arithmetic calculator for performing mathematical operations. Use when the user needs to perform calculations, arithmetic operations, or mathematical computations including addition, subtraction, multiplication, division, exponentiation, and more complex mathematical expressions.

kaualealz Bundle

---

lilialla Bundle

面向法律文书格式处理的 Agent Skill，把用户提供的 DOCX 模板作为格式事实来源，支持自然语言模板执行、内容锁定、DOCX 结构审计和发布档强制 LibreOffice + Poppler 视觉校验。

lang7910 Skill

Build, fill, audit, and derive logistics-simulation availability, cycle-time, safety-zone, input, and theoretical-capacity workbooks. Use when Codex needs to work with TA/availability calculation tables, station equipment counts, multilingual process PPT/PDF/Word evidence, Gantt-style cycle-time sheets, safety-zone or lockout-zone summaries, simulation input workbooks, model-variant cycle times, plant/manual loss factors, or theoretical line-capacity comparisons for manufacturing or logistics simulation across automotive body-in-white/welding or other production/logistics domains.

lang7910 Skill

Core Plant Simulation logistics modeling skill skeleton. Use when Codex needs shared project conventions, modeling assumptions, MCP tool usage patterns, or future reusable workflow guidance for Plant Simulation logistics simulation work.

xiaoyiweio Bundle

Preflight security scanner for AI coding agents — scans deployment config, skills/MCP servers, memory/sessions, and AI agent config files (hooks injection) for secrets, PII, prompt injection, and dangerous patterns. Runs 4 model behavior probes (persuasion, sandbagging, deception, hallucination). Supports LLM-enhanced semantic analysis. Works with OpenClaw, Claude Code, Cursor, Codex, and OpenCode. Use when a user asks for a security audit, health check, or wants to scan their AI agent setup for vulnerabilities.

finger88 Bundle

Launches a multi-agent debate on any topic with real-time evidence verification and causal reasoning chains. Use this skill when the user says "debate", "start a debate", "run a debate on", "argue about", "red team this", "analyze from multiple perspectives", or provides a topic for critical examination. Supports balanced and red-team modes, configurable depth and rounds, domain-aware evidence gathering, and generates structured Markdown reports with conclusion profiles and 24h watchlists.

finger88 Bundle

Constructs a complete structured debate turn with 5-element reasoning chains, rebuttals, and evidence references. Use this skill when a debater agent needs to construct an argument, build a debate turn, generate rebuttals against opponent, respond to mandatory judge response points, create a structured argument with evidence and causal reasoning chain, or produce a debate round output with historical wisdom and speculative scenarios.

finger88 Bundle

Performs independent verification, causal chain audit, analogy validation, and generates structured rulings for debate rounds. Use this skill when the judge agent needs to audit a debate round, verify claims independently, check causal validity of arguments, produce a structured ruling, identify mandatory response points for next round, perform independent second-pass verification, evaluate reasoning quality, assess historical wisdom references, or review speculative scenarios.

somnus-j-307 Bundle

Converts casual,口语化 instructions into structured,工程化 prompts that maintain original intent while improving effectiveness. Use when user says "转换提示词", "优化prompt", "帮我改一下这个指令", "让这个prompt更有效", "prompt优化", "prompt转换", "帮我写个更好的prompt", "这个指令怎么写更好", "把我的想法变成prompt", "prompt engineering", or any request to improve/transform a prompt or instruction.

kohosoda Bundle

Track and evaluate recent AI model, coding-agent, Codex, Claude Code, and developer-tool updates. Use when Codex is asked to catch up on AI news, summarize release notes, compare new model or agent capabilities, monitor changelogs, produce a weekly AI update digest, or recommend changes to skills, hooks, automations, model choices, or local AI coding workflows based on recent updates.

finger88 Bundle

Manages the claim state machine for a multi-agent debate system. Use this skill when the orchestrator needs to update the claim ledger, record new claims from a debate turn, change claim status based on judge ruling, track claim state transitions, extract claims from arguments, perform batch updates from judge rulings, or manage the claim lifecycle with audit trail persistence.

leima-max Bundle

Origin Pro MCP Skill / Origin Pro MCP 技能