Core

DeepEval — LLM evaluation framework, RAG metrics, hallucination detection, red-teaming, CI/CD integration

ByteDance's open-source super agent harness — spawns parallel sub-agents, Docker sandbox execution, persistent long-term memory, modular skills (research/report/slides). Built on LangChain + LangGraph. Triggers on: 'deerflow', 'deer-flow', 'bytedance agent', 'super agent harness', 'LangGraph agent orchestration', 'multi-hour agent tasks', 'spawn sub-agents', 'agent sandbox docker', 'persistent agent memory', 'agent skills system', 'AI research orchestrator', 'long-running agent tasks', 'agent with memory', 'langgraph orchestrator', 'IM channel agent integration'.

Context compression for YAMTAM — nén JSON/structured tool output trước khi vào LLM. Hiệu quả với JSON (50-72% tiết kiệm); text thuần cần bản [all].

LLM observability with Langfuse — tracing, evals, prompt management, cost tracking

Outlines — structured generation with guaranteed JSON/regex output from local LLMs

YAGNI-first code-minimalism ladder — before writing code, check in order: does this need to exist? already in the codebase? stdlib? native platform feature? installed dependency? one line? Cuts unnecessary code/tokens while never trading away validation, error handling, security, or accessibility. Triggers on: 'ponytail mode', 'write minimal code', 'don't over-engineer this', 'yagni', 'is there already something for this', 'use native instead of a library', 'keep this simple', 'fewest lines possible', 'lazy senior dev mode'.

Full-cycle AI coding skill: plans, builds, tests, lints, fixes bugs, and writes production-grade docs. Auto-detects project state and adapts its 8-phase pipeline.

Deploy web apps with backend APIs, database, and file storage. Use when the user asks to deploy or publish a website or web app and wants a public URL. Uses HTTP API via curl.

Use this skill to measure performance baselines, detect regressions before/after PRs, and compare stack alternatives.

Turn a one-line objective into a step-by-step construction plan for multi-session, multi-agent engineering projects. Each step has a self-contained context brief so a fresh agent can execute it cold. Includes adversarial review gate, dependency graph, parallel step detection, anti-pattern catalog, and plan mutation protocol. TRIGGER when: user requests a plan, blueprint, or roadmap for a complex multi-PR task, or describes work that needs multiple sessions. DO NOT TRIGGER when: task is completable in a single PR or fewer than 3 tool calls, or user says "just do it".

Query the Yana AI codebase knowledge graph via codebase-memory-mcp. Use instead of grep/glob when exploring call chains, finding callers/callees, understanding architecture, or tracing impact of changes. Triggers on: 'who calls X', 'trace path', 'find callers', 'search graph', 'cbm', 'knowledge graph', 'what calls', 'call chain', 'what uses', 'where is X defined', 'architecture overview', 'impact of changing'.

Use when the user asks to create a CodeTour .tour file — persona-targeted, step-by-step walkthroughs that link to real files and line numbers. Trigger for: create a tour, onboarding tour, architecture tour, PR review tour, explain how X works, vibe check, RCA tour, contributor guide, or any structured code walkthrough request.

codexmate

Multi-platform content distribution across X, LinkedIn, Threads, and Bluesky. Adapts content per platform using content-engine patterns. Never posts identical content cross-platform. Use when the user wants to distribute content across social platforms.

Navigation guide cho ECC — onboarding, tìm skills/agents/commands phù hợp, setup, troubleshooting. Dùng khi cần hỏi 'làm X với ECC như thế nào?'

Evidence-first mailbox triage, drafting, send verification, and sent-mail-safe follow-up workflow for ECC. Use when the user wants to organize email, draft or send through the real mail surface, or prove what landed in Sent.

Scrape and crawl websites for AI with Firecrawl — scrape single URLs to clean Markdown/HTML, crawl entire sites with depth/path filters, extract structured data with LLM schema, use map to discover all URLs, and batch scrape multiple pages in parallel.

Fact-forcing gate that blocks Edit/Write/Bash (including MultiEdit) and demands concrete investigation (importers, data schemas, user instruction) before allowing the action. Measurably improves output quality by +2.25 points vs ungated agents.

Use when generating elite GSAP-heavy frontend pages with strict AIDA structure, wide hero typography, and gapless bento grids.

Use when building stateful multi-step agents, agent graphs, or workflows with LLMs. Triggers on: 'langgraph', 'state graph', 'stateful agent', 'agent workflow', 'agent loop', 'multi-step agent', 'persistent agent', 'human-in-the-loop agent', 'agent with memory', 'graph-based agent'.

Session-end memory garbage collector. Promotes valuable L2 session facts to L1 atomic memory, wipes L2, and rotates oversized audit logs. Run at end of session to prevent context inflation and storage bloat.

Design for mobile — touch targets, thumb zones, gesture patterns, iOS and Material 3 platform conventions, mobile form design, and responsive breakpoints. Use when building mobile-first UI, reviewing a layout for phone/tablet, asked about "touch targets", "thumb zone", "iOS conventions", "Android Material", "mobile form", or "responsive design". Do NOT use for desktop-only UIs — use `ux-heuristics` for general usability.

Production-ready UI motion system for React/Next.js. Use when implementing animations, transitions, or motion patterns.

Cross-stack source code asset audit — classifies every file, detects embedded third-party libraries, and delivers actionable four-level verdicts per module with interactive HTML reports.

Detect and eliminate predictable AI writing tells from prose. Use when the user asks to "make this sound less like AI", "remove AI tells", "make it more human", "this sounds robotic", "clean up the writing", "too formal", "sounds generated", or requests a writing/copy review. Do NOT use for code — this skill targets prose, copy, and documentation only.

Recognize and extract text from images or PDFs using Surya OCR (datalab-to/surya). Use when a user uploads a document, screenshot, scan, or photo and wants the text extracted or analyzed. Surya supports 90+ languages and works on both images and PDFs without an internet connection.

Build full-stack web applications with SvelteKit — file-based routing, SSR, SSG, API routes, and form actions in one framework.

Token budget analytics and ROI scoring for AI agent actions. Track token cost per fix, detect wasteful loops, auto-route to fast tier when burn rate is too high. Integrates with token-budget-guard.sh hook. Triggered by /cost-report, "how much did that cost", "token usage", "am I wasting tokens".

Opinionated, evolving constraints to guide agents when building interfaces

Design all 7 UI states for any data-dependent view: empty, loading, skeleton, partial, complete, error, offline. Use when asked about "loading state", "skeleton screen", "error message", "empty state", "optimistic UI", "what happens while data loads", or before marking any data-dependent view as done. Every view that fetches data must pass all 7 states before ship.

Use when the user has UI screenshots or design exports that need batch conversion into Vue 3 components, especially with Vant, Element Plus, or Ant Design Vue.

Head-to-head comparison of coding agents (Claude Code, Aider, Codex, etc.) on custom tasks with pass rate, cost, time, and consistency metrics

Build an evidence-backed ECC install plan for a specific repo by sorting skills, commands, rules, hooks, and extras into DAILY vs LIBRARY buckets using parallel repo-aware review passes. Use when ECC should be trimmed to what a project actually needs instead of loading the full bundle.

Skill for discovering and researching autonomous AI agents, tools, and ecosystems using the AgentFolio directory.

Build persistent multi-agent operating systems on Claude Code. Covers kernel architecture, specialist agents, slash commands, file-based memory, scheduled automation, and state management without external databases.

Every product will be AI-powered. The question is whether you'll build it right or ship a demo that falls apart in production.

REST API design patterns including resource naming, status codes, pagination, filtering, error responses, versioning, and rate limiting for production APIs.

Use this skill to automate visual testing and UI interaction verification using browser automation after deploying features.

Đọc Gmail, báo cáo thư chưa đọc. Dùng khi muốn check inbox nhanh mà không cần mở browser.

This skill embodies the principles of "Clean Code" by Robert C. Martin (Uncle Bob). Use it to transform "code that works" into "code that is clean."

Django testing strategies with pytest-django, TDD methodology, factory_boy, mocking, coverage, and testing Django REST Framework APIs.

Build evaluation frameworks for agent systems. Use when testing agent performance systematically, validating context engineering choices, or measuring improvements over time.

Neural search via Exa MCP for web, code, and company research. Use when the user needs web search, code examples, company intel, people lookup, or AI-powered deep research with Exa's neural search engine.

Verify fix commits address audit findings without new bugs

GitHub repository operations, automation, and management. Issue triage, PR management, CI/CD operations, release management, and security monitoring using the gh CLI. Use when the user wants to manage GitHub issues, PRs, CI status, releases, contributors, stale items, or any GitHub operational task beyond simple git commands.

Eliminate generic-looking UI — anti-default checklist, custom hover states, precision typography, non-stock spacing, surface depth, and production-grade polish that separates professional from template. Use when asked to "make this look less generic", "UI looks Bootstrap-y", "looks like a template", "add polish", "impeccable UI", "production quality design", "make it look custom", "it looks like every other SaaS app", "remove default styles", or "design lacks personality". Do NOT use for: design system token architecture — see design-system-gen. Do NOT use for: animation timing specifics — see motion-design.

Review L2 session facts and promote valuable ones to L1 Atomic Memory. Use at end of session or when L2 accumulates facts worth keeping permanently. Runs add-fact.sh for each promoted fact.

Scan codebase and git history for leaked secrets — API keys, tokens, passwords, private keys. Uses whispers-style regex patterns + gitleaks. Run as /leak-check before any git push. Alerts on Claude API keys, GitHub tokens, AWS credentials, and generic high-entropy strings.

AI-powered Claude Code skill that performs deep code review using formal logic and reasoning frameworks to detect bugs, anti-patterns, and security risks beyond what linters catch.

Convert mọi file (PDF, Word, Excel, PowerPoint, Image, Audio, HTML, CSV, YouTube URL) thành Markdown cho LLM. Tích hợp vào YAMTAM vault để import tài liệu.

Self-hosted open-source collaboration platform — chat, bots, webhooks, plugins, AI integration, Slack/Teams alternative with full data control. Triggers on: 'mattermost', 'self-hosted slack alternative', 'open source teams alternative', 'mattermost bot', 'mattermost webhook', 'mattermost plugin', 'mattermost API', 'mattermost integration', 'AI agent mattermost', 'deploy mattermost', 'mattermost docker', 'mattermost kubernetes', 'enterprise chat self-hosted', 'mattermost AI features', 'mattermost incoming webhook', 'mattermost slash command'.

Use before touching any code when user gives a task, spec, or feature request. Forces writing a plan before implementation. Examples: 'implement X', 'add feature Y', 'fix Z and also update A', any multi-step task that would touch 2+ files.

Design and implement SLIs, SLOs, and error budgets for production services. Covers SLI selection, SLO target-setting, error budget calculation, burn-rate alerting, and error budget policy. Use when asked about "SLO", "SLI", "error budget", "reliability targets", "burn rate", "uptime SLA", or when designing alerting for a production service. Do NOT use for: application-level performance optimization — use `web-performance` for frontend.

Build lightweight AI agents with HuggingFace Smolagents — use CodeAgent (writes Python to act) or ToolCallingAgent (JSON tool calls), add built-in or custom Tools, orchestrate multi-agent pipelines with ManagedAgent, and run locally or via HF Inference API.

Three-phase vulnerability chain analysis — parallel agents find individual weaknesses, then a synthesis step identifies which combinations escalate to critical impact. Inspired by Strix (usestrix/strix) "Graph of Agents" pentesting model.

Use when provisioning a headless Windows 11 VM in Docker for testing, running Windows-only tools, or installing Claude Code on Windows. Triggers on: 'windows vm', 'windows docker', 'headless windows', 'windows 11 container', 'windows vm ssh', 'kvm windows', 'dockur windows', 'windows dev environment', 'windows claude code', 'windows test environment'.

Write and maintain Architecture Decision Records (ADRs) — when to write one, the standard format, status lifecycle, how to link related decisions, and how to surface ADRs in a codebase. Use when asked to "write an ADR", "document this decision", "architecture decision record", "why did we choose X over Y", or "we need a record of this choice". Do NOT use for: general documentation writing — ADRs are specifically for significant architecture decisions.

Use when an agent needs to read the internet without paid API keys — Twitter/X, Reddit, YouTube, Bilibili, GitHub, TikTok, Xiaohongshu, RSS, web pages. Triggers on: 'agent-reach', 'agent đọc internet', 'đọc Twitter không API', 'đọc Reddit không API', 'agent xem YouTube', 'agent đọc Bilibili', 'agent đọc B站', 'agent đọc XHS', 'agent đọc Xiaohongshu', 'read social media free', 'agent internet eyes', 'no API key scraping', 'unified social reader'.

Establish CSS/Tailwind baseline quality — font smoothing, line-height defaults, box-sizing, focus rings, color contrast, Tailwind anti-patterns, and CSS containment. Use when asked to "set up CSS baseline", "Tailwind best practices", "fix font rendering", "global CSS foundation", "typography defaults", "CSS anti-patterns", "improve CSS quality", "line-height reset", "normalize styles", or reviewing a new project's base styles for correctness. Do NOT use for: full design system token architecture — see design-system-gen. Do NOT use for: component-level animation — see motion-design.

Build a source-derived writing style profile from real posts, essays, launch notes, docs, or site copy, then reuse that profile across content, outreach, and social workflows. Use when the user wants voice consistency without generic AI writing tropes.