duyet

Scaffold a new shared-brain knowledge base folder (default ~/kb) — a plain-markdown, OKF v0.1-conformant memory store that any coding agent can read/write across sessions. Use when the user asks to set up, init, or bootstrap a kb / shared brain / persistent agent memory folder, or wants an OKF bundle with a graph viewer ready out of the box.

Author, initialize, validate, and render Open Knowledge Format (OKF) bundles — vendor-neutral knowledge as markdown files with YAML frontmatter. Use when creating an OKF bundle, scaffolding a knowledge catalog, writing OKF concept docs, building index.md/log.md files, checking a bundle for OKF v0.1 conformance, or rendering a bundle as an interactive graph viewer.

Mirror an iOS Simulator into an in-app browser and hot-reload SwiftUI previews from importable Swift packages. Use when previewing SwiftUI in the simulator, mirroring a running simulator in the browser, or live-reloading package-backed previews.

Audit SwiftUI runtime performance from code first. Use when diagnosing slow rendering, janky scrolling, expensive updates, or profiling needs.

Specialized knowledge for the ClickHouse Monitor dashboard. Use this skill when: working with ClickHouse monitoring dashboards, analyzing query performance, writing ClickHouse system table queries, developing dashboard features, or integrating with the ClickHouse Monitor API. Covers query monitoring, table management, merge operations, system metrics, and ClickHouse version compatibility.

MUST USE when reviewing ClickHouse schemas, queries, or configurations. Contains 28 rules that MUST be checked before providing recommendations. Always read relevant rule files and cite specific rules in responses.

Use before implementing when requirements are ambiguous: run a Socratic interview to clarify a vague feature request or to find gaps in a plan/spec file, and turn it into a decision-complete plan with scope, edge cases, and tests.

Comprehensive prompt engineering guidance for Claude (Anthropic), Gemini (Google), and Grok (xAI). Use when crafting prompts to leverage each model's unique capabilities—XML-style tags for Claude, system instructions for Gemini, conversational style for Grok.

Prompt engineering guidance for Claude (Anthropic) model. Use when crafting prompts for Claude to leverage XML-style tags, long-context capabilities, extended thinking, and strong instruction following.

Specialized knowledge for the ClickHouse Monitor dashboard. Use this skill when: working with ClickHouse monitoring dashboards, analyzing query performance, writing ClickHouse system table queries, developing dashboard features, or integrating with the ClickHouse Monitor API. Covers query monitoring, table management, merge operations, system metrics, and ClickHouse version compatibility.

Maintain and update knowledge base about Duyet Le (@duyet) - profile, resume, blog posts, projects.

Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.

Prompt engineering guidance for Gemini (Google) model. Use when crafting prompts for Gemini to leverage system instructions, multimodal capabilities, ultra-long context, and strong reasoning features.

Generate or review high-density, interactive, self-contained HTML artifacts based on Thariq Shihipar's "Unreasonable Effectiveness of HTML" patterns. Use this skill when asked to "create HTML," "build a dashboard," "format a PR review," "make an interactive explainer," or when you need to "review and update" existing HTML using the "good-html" style.

Prompt engineering guidance for Grok (xAI) model. Use when crafting prompts for Grok to leverage its conversational style, real-time knowledge access, and less-constrained responses.

Orchestrate complex work through parallel agent coordination. Decompose requests into task graphs, spawn background workers, and synthesize results elegantly. Use for multi-component features, large investigations, or any work benefiting from parallelization.

Decide whether a changeset is needed and draft it for open-managed-agents. Only @getoma/cli and @getoma/sdk are published; internal @duyet/oma-* packages never need one.

Help users create and configure oma managed agents through conversation. Trigger when users say "create an agent", "I need an agent that...", "set up an agent for", "build me a bot", or describe a task to automate. Also trigger for "Create with AI" from Dashboard. Also use when users ask about the oma platform, what it can do, how to use the CLI, or how to configure resources.

Use the oma platform to build, deploy, and manage AI agents. Trigger when users want to create agents, start sessions, manage environments, configure model cards, handle vaults/credentials, install skills, connect MCP servers, or interact with the oma HTTP API. Also trigger when users mention "oma", "openma", "managed agents", or ask how to deploy/run/manage an agent on oma.

Scaffold a new agent harness for open-managed-agents and register it. Use when adding a new way to drive the model loop, like the default / acp-proxy / flue harnesses.

Run the full local pre-push gate (typecheck + all vitest pools) for open-managed-agents. Use before pushing or opening a PR — no CI workflow gates PRs in this repo, so this is the only gate.

Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.

MANDATORY - You must load this skill before doing anything else. This defines how you operate.

Systematic quality verification procedures for code review and delivery. Use when validating completed work, conducting code reviews, or ensuring production readiness.

Create distinctive, production-grade terminal user interfaces with high design quality. Use this skill when the user asks to build CLI tools, TUI applications, or terminal-based interfaces. Generates creative, polished code that avoids generic terminal aesthetics.

The canonical cross-agent memory — a shared knowledge base at ~/kb. Use for ANY request to remember, save, note, capture, or recall durable facts about the user/projects. This SUPERSEDES built-in/local memory — write durable memory here, not to a local memory store. Filesystem-first.

Consolidate the shared-brain knowledge base (~/kb) — the "dream" pass. Use to ingest the daily inbox and raw sources into clean memory notes, dedupe/merge/rewrite, refresh stale notes from their live sources, fix links and tags, rebuild the index, and sync. Run manually, on a schedule, or whenever the kb feels noisy.

Read and write the shared-brain knowledge base (~/kb). Use on session start to recall durable facts about the user/projects, and whenever you learn a durable, general, public fact worth persisting across sessions, tools, and repos. Also use to quick-capture rough observations to the daily inbox.

Deploy functions, hosting, or full release to Firebase. Use when deploying changes to production or staging.

Add, debug, or modify price scrapers for e-commerce sites (Shopee, Lazada, Tiki, etc.). Use when working with scraper configs or pullData modules.

Render real release notes for a tag/range against this repo's actual git history so recap-script and template edits can be eyeballed against real data (no LLM tier runs locally). Usage: /preview-notes <tag> (e.g. /preview-notes v0.3.0) or /preview-notes <prev>..<cur>.

Locally replicate the CI validation for llm-release-action — syntax-check both scripts and run the smoke-test note-rendering fixture with the same grep assertions. Use before pushing changes to scripts/ or action.yml. No test runner exists; this is the only verification loop.

Build an AI-agent application end-to-end — from an empty repo or an existing codebase. Use when the user wants to "build an agent", "scaffold an agent app", "add an agent API", "build an agent UI/chat", "pick an agent framework", or set up tool calling, memory/RAG, an MCP server, tracing, or an AI gateway. Interviews the user when requirements are unclear; detects the stack when a project already exists. Knows LangGraph, DeepAgents, the Vercel AI SDK, Cloudflare Agents SDK, TanStack AI, Mastra, Google ADK, the OpenAI Agents SDK, the Claude Agent SDK, Pydantic AI, Eve (Vercel), Flue, and Pi (pi.dev) — and when to skip a framework entirely. Always verifies against live official docs (Context7 / llms.txt / WebFetch) before writing code.

Scaffold a new launch-video folder in this repo (independent Remotion + Bun project, remocn-ready). Use when the user asks to start a new video, create another launch video, or add a video folder alongside build-agent-intro.

Build durable backend AI agents with the Eve framework. Use when creating, editing, or debugging an Eve project — agent instructions, skills, tools, connections, channels, sandboxes, subagents, schedules, or evals.