AI & ML

zap-coding-agent Skill

Use when building intuition for how LLMs work — attention mechanisms, tokenization gotchas, why transformers work, context window behavior, temperature/sampling, prompt engineering from first principles, or when debugging why a model gives unexpected outputs. Applies Karpathy's "understand the math, not just the API" approach.

zap-coding-agent Skill

Use when normalizing Splunk fields to standard names so searches, dashboards, and detections work across data sources — mapping vendor fields to CIM (Common Information Model), building/validating data models, tagging events, or when a correlation search or Enterprise Security panel returns nothing because fields don't match. Covers CIM data models, field aliases, tags, eventtypes, and acceleration.

cuimuxuan Bundle

Coordinate an end-to-end academic paper, thesis, dissertation, review paper, or manuscript workflow by routing work across research verification, PDF/DOCX parsing, writing, academic figures, de-AI polishing, post-draft review, and formatting skills. Use when the user asks to manage a full paper project, plan a thesis from source materials to final DOCX, decide which academic skill should handle a task, run gated human-AI paper production, coordinate evidence gaps, field confirmation, reviewer-comment revision planning, final polish handoffs, or default multi-agent/sub-agent parallel review, verification, QA, and action-plan decomposition for substantial paper workflows. Chinese triggers: 论文全流程, 毕业论文, 学位论文, 综述论文, 论文总控, 编排论文工作流, 多agent论文流程, 子agent并行审查, 多智能体论文审查, 并行论文核验, 待补证据流程, 评审意见流程, 最终润色流程.

cuimuxuan Bundle

Post-draft academic manuscript review against benchmark papers and editorial standards. Use only after an initial full manuscript draft, thesis chapter draft, or journal paper draft exists and the user asks for review, grading, gap analysis, target-paper comparison, SCI/Q1 or Applied Energy-style readiness assessment, desk-reject risk review, page/paragraph/sentence-level audit, benchmark-literature comparison, reviewer-comment response planning, next-version optimization planning, or multi-agent/sub-agent parallel review across scope fit, novelty, scientific logic, evidence support, language audit, citation proximity, figure/format QA, and reviewer-comment action planning. Do not use during outline planning, literature search, figure generation, citation verification, formatting, or de-AI polishing unless a complete draft exists. Chinese triggers: 初版论文后评审, 论文生成后评审, 对标文献差距, SCI一区评审, 论文质量评判, 下一版优化方案, 与对标论文比较, 审稿意见处理, 多agent论文评审, 子agent并行评审, 多智能体稿件审查, 期刊匹配度审查, 创新性审查, 科学逻辑审查.

sanjai52 Skill

Runs after agent-strategy outputs PROCEED: yes. Produces a written implementation plan before touching any file. Includes files to create/modify, approach, layer compliance, Zod schema plan, migration number if needed, and tests required.

chu459 Bundle

Control Claude Code as an external sub-agent through the local cc-orchestrator MCP/CLI and CCSwitch profiles, with .agent-workspace governance for logs, reports, dashboards, archives, rollback notes, templates, policies, and MCP path repair. Use when the user asks Codex to use Claude Code, CCSwitch, external model routing, visible Claude Code windows, ClaudeCode as subagents, MCP control of Claude Code, configurable multi-agent workflows, workspace/status/cleanup/migration/archive of Claude Code run artifacts, repair .mcp.json paths, or folder policies where Codex is the controller and Claude Code agents are workers.

thomasblc Bundle

SkillsIndex MCP

contentscoin Skill

QR Agent Studio를 Claude Code·Codex·ChatGPT에 연결하는 방법을 클라이언트별로 한 곳에서 안내하는 스킬.

vari-arti Skill

Create or update the reused.md artifact for an OpenSpec change — audit existing code for functionality to reuse, extend, or refactor before writing tasks. Use when planning implementation, after data-model and before tasks, or when the user mentions code reuse, DRY, duplicate logic, or extending existing components in an OpenSpec change.

vari-arti Skill

Create or update the data-model.md artifact for an OpenSpec change — DB structure, relationships, migrations, and storage conventions. Use when planning persistent data changes, after design and before tasks, or when the user mentions data model, schema, migrations, or database structure in an OpenSpec change.

interfluve-wav Bundle

Behavioral guidelines to reduce common LLM coding mistakes. Derived from Andrej Karpathy's observations. Use ALWAYS when writing, reviewing, or refactoring code. Covers assumption surfacing, simplicity, surgical edits, and goal-driven execution with verification.

ffoysal Bundle

An interactive lesson that teaches the Model Context Protocol (MCP) by building a real, working MCP server from scratch. Use this skill whenever the user wants to learn MCP, understand MCP concepts (host, client, server, tools, resources, prompts, transport), asks "what is MCP" or "how does MCP work", wants to build an MCP server, mentions building a database-backed or HTTP MCP server, or asks for an MCP tutorial, walkthrough, or hands-on guide. Trigger this even if the user only says something like "teach me MCP", "I want to build an MCP server", or "explain MCP tools vs resources" — they want this guided lesson, not a one-line answer.

contentscoin Bundle

QR Agent Studio API, MCP, Agent Skill, OpenAPI plugin 연동을 설계하고 구현 계획을 작성하는 개발/설계형 스킬.

qiyang-job Bundle

当项目进入多轮对话、跨文件开发、暂停后恢复、多人或多 agent 协作，或需要让 AI 在干活时持续记住项目全貌、关键规则、近期状态与历史教训时使用。适用于“先同步项目上下文再继续”“恢复昨天的开发状态”“做完后更新交接与记忆文件”“避免重复犯同样的问题”等场景。

sunwoongkyu Bundle

Claude Code 스킬 제조 공장 — 막연한 요청을 10라운드 + α 문답으로 충분히 기획하고, 공개 저장소·OSS·사내 자산에서 부품을 발굴·안전성진단·분해하여 검증된 부품만 적용한 뒤, QC로 출하검사한 완성 SKILL.md를 제조한다. 단일 책임·자기완결·500줄 권장 — 거대 스킬을 만들지 않는다. 단일/배치 2가지 제조 모드 지원 (포털은 인프라 필요로 에이전트 전용). 사용자가 '스킬 만들어', '스킬 제작', 'skill-create', 'skill-create-코어5', '~하는 스킬 만들어줘'를 요청할 때 사용. (LLM 의존형 에이전트 제작은 /에신-llm-dependent-agent-create — 형제 공장 별칭 '에신')

avizmarlon Skill

Use Notion's REST API directly for batch database operations (querying, updating many rows, pagination, file uploads, creating databases). Auto-load when tasks mention Notion AND batch operations like "database", "update many", "iterate pages", "audit", "scan", or touch ≥3 rows. Prefer MCP (Notion search/create) for single-page CRUD only.

69mannying Bundle

ML research agent persona and operating protocols. Use when designing experiments, choosing problems, reading literature, analyzing training runs, writing research notes, debugging models, or when the user invokes research mode. Embodies outcome-first reasoning, fast iteration loops, taste-building, and honest logging over trend-following.

nexenio Skill

Build a decision-tree "navigator" skill over a subset of installed skills. Embeds each skill's description, clusters them bottom-up into a 4-tier tree (Wing > Room > Zone > Station > leaf), LLM-labels every node, and writes one decision-node SKILL.md per node plus a single root entry skill. An agent then enters the root, answers a routing question, and follows branch hints down tier by tier until it reaches the right leaf skill. Use when a skills directory has too many skills to route by description alone and you want one navigable entry point. NOT for creating new skills, installing skills, or grouping 2-3 skills (use a plain router for that).

mingze21 Bundle

八字 + 紫微斗数 AI 排盘与综合分析。当用户提供生辰（阳历/农历日期、时辰、性别）询问八字、紫微、命理、命盘、流年大运相关问题时使用。基于 Yiqi 算法库 + enrichBazi 补全层精准排盘（杜绝 LLM 自行排盘出错），支持按需独立分析八字 / 独立分析紫微 / 两盘交叉印证。

avizmarlon Skill

Select the right LLM for a task by consulting benchmarks and real performance data. Use when choosing between models (chatbot, agent, text classification, code generation, etc.), comparing model performance, or validating model availability and pricing.

avizmarlon Skill

Maintain configuration and capability parity across multiple AI agent runtimes (Claude, Codex, Cursor, Gemini, custom agents). Use when syncing rules, skills, MCPs, configs, or credentials across team AI tools, or when debugging drift between agents.

avizmarlon Skill

Use Notion REST API (not MCP) for batch operations (≥3 rows). Applies to database queries, batch updates, scans, pagination, file uploads, database creation, and page replication. MCP is acceptable only for single-page CRUD operations.

avizmarlon Skill

Hooks as enforcement mechanism for AI behavior — use when a written rule fails repeatedly in generation flow and deterministic structural enforcement is needed. Triggers include detecting cyclic rule failure, asking about Claude Code hooks/lifecycle events (SessionStart, UserPromptSubmit, PreToolUse, PostToolUse, Stop), or when a behavior guard needs to execute outside the AI's inference loop.

avizmarlon Skill

Enforce hook lifecycle synchronization across all AI agent tools. When creating, modifying, or removing hooks in any AI tool (Claude Code, Codex CLI, Cursor, Gemini, Kimi, etc.), propagate changes to all equivalent tools in the same session. Use when implementing cross-tool runtime enforcement.

topease-ai Bundle

通过 TOPEASE/Tradee Customs MCP 查询和分析全球海关贸易记录。用于用户要求查询进出口海关数据、买家或供应商出货记录、产品关键词贸易数据、HS 编码贸易数据、国家或市场进出口记录、进口商/出口商线索，或要求配置、接入、排查 TOPEASE 海关数据 MCP 服务时。

avizmarlon Skill

Permanent learning — AI agent proposes learnings from sessions, user approves in chat, integrations saved to auto-loaded files. Triggers on session end or manual /learn command. Use when consolidating discoveries, user says learn/consolidate/document, or end-of-session signals.

hinkok Bundle

Multi-model decision workflow for Hermes: Conductor plans and synthesizes, Worker does high-volume work, and Critic red-teams the result. Use for important strategy, research, content, niche, product, business, and high-stakes decisions; triggers include троица, используй троицу, через 3 модели, GPT + DeepSeek + Gemini, council, red team this, plan-execute-critique.

tientruongminh Bundle

Jupyter MCP Server Skill

saranambiar Bundle

Integrate raw product demo recordings into agent-created videos, including crop decisions, zoom preservation, camera swipe transitions, UI readability checks, and timing around narration. Use when a user provides screen recordings, product walkthrough clips, SaaS demos, app footage, browser recordings, or asks to transition into/out of product demo footage.

saranambiar Skill

End-to-end video production workflow for coding agents using HyperFrames, FFmpeg, captions, audio sync, music beds, and render QA. Use when a user asks for a complete video, multi-scene edit, launch video, product/demo video, UGC-style edit, enterprise explainer, or final assembled MP4 from prompt/script/assets.

saranambiar Skill

Convert rough video prompts, scripts, reference images, and edit notes into decision-complete scene specs and storyboards for agent-led video production. Use when a user describes a scene, gives timestamped narration, asks to preserve previous scenes, references a concept image, or requests a multi-scene video plan.

ffffj-ai Bundle

Self-onboard an LLM agent to the ACPrompt network — STEP 1 self-audit your runtime, STEP 2 connect via the method that fits (paste-link / OAuth / raw token / install command), then register, heartbeat, exchange Layer 1/2 messages, collaborate on cross-owner projects, propose modules, file disputes, claim open tasks, and self-integrate any framework — without an SDK. Compatible with Claude Skills (SKILL.md) loading convention.

fandry96 Bundle

Autonomous HestiaCP server management with AI agents. Two-brain Orchestrator+Sentinel architecture for VPS monitoring, troubleshooting, security hardening, PHP-FPM tuning, Exim mail forensics, and post-compromise detection. Use when: HestiaCP, VPS, server monitoring, DevOps agent, Linux admin, Nginx/Apache, PHP-FPM, Exim, Fail2Ban, ClamAV, SSL, backup.

fandry96 Skill

Gemini Embedding 2 — Google's natively multimodal embedding model (GA April 2026). Maps text, images, video, audio, and PDFs into a unified 3072-dim vector space with MRL support (truncate to 128-3072). Use when: embeddings, semantic search, RAG, clustering, classification, multimodal retrieval, vector database, MRL truncation, cosine similarity, narrative memory, code retrieval, cross-modal search, agentic RAG.

fandry96 Skill

Specialized agent for architecting Firebase Data Connect layers. Use for "generating GraphQL schemas," "optimizing PostgreSQL/AlloyDB queries," and "configuring IAM-backed database connectors."

clawrun-sh Bundle

Use when the user needs to interact with GitHub — checking CI status, viewing PRs, listing issues, querying the API, or any GitHub-related task. Triggers include "check CI", "list PRs", "view issue", "GitHub API", or any request involving repositories, pull requests, issues, or workflows.