practical-stack

practical-stack/ai-lab

Command Claude Code

Analyze task description and generate appropriate git branch name.

practical-stack/ai-lab

Command Claude Code

Create structured learning content from research materials.

practical-stack/ai-lab

Command Claude Code

Diagnose LLM component type (Skill/Agent/Command) for a feature → create or update.

practical-stack/ai-lab

Command Claude Code

Create or update a draft PR with Summary and Test sections.

practical-stack/ai-lab

Skill Claude CodeCodex

Generate and validate YAML frontmatter for docs folder documents. Use when users want to add frontmatter, update frontmatter, generate frontmatter metadata, or validate existing frontmatter against the schema.

practical-stack/ai-lab

Skill Claude CodeCodex

Transform research materials into structured learning content with bilingual support. Use when users want to create learning content from research, convert research to learning paths, make learning materials, or translate learning content between languages.

practical-stack/ai-lab

Skill Claude CodeCodex

Deep analysis of LLM agent and plugin repositories to extract architecture, patterns, and insights. Use when users want to analyze an LLM repo, extract patterns from a repository, do a repository deep dive, or understand an AI agent codebase. Do NOT use for general code review, simple file exploration, or non-LLM…

practical-stack/ai-lab

Skill Claude CodeCodex

Guide for creating and validating specialized AI agents (subagents). Use when users want to create a new agent, make a subagent, define agent configurations, validate agents, or need guidance on agent architecture. Do NOT use for creating skills (use meta-skill) or commands (use meta-command).

practical-stack/ai-lab

Skill Claude CodeCodex

Guide for creating and validating Claude Code slash commands. Use when users want to create a command, define command specs, configure allowed-tools or argument-hint fields, or validate and audit existing commands. Do NOT use for creating skills or agents (use meta-skill, meta-agent).

practical-stack/ai-lab

Skill Claude CodeCodex

Diagnose features into the right LLM component type — Skill, Agent, or Command wrapper. Use when users ask whether something should be a skill, need to decide between command or agent, want to diagnose a feature's type, or make structure and architecture decisions for LLM components. Do NOT use for actual…

practical-stack/ai-lab

Skill Claude CodeCodex

Generate high-quality prompts using proven prompt engineering techniques. Use when users want to write a prompt, create or improve prompts, work on meta-prompts, or need to craft effective instructions for AI tasks. Do NOT use for simply asking questions, executing existing prompts, or non-prompt-related tasks.

practical-stack/ai-lab

Skill Claude CodeCodex

Extract reusable patterns from completed session work. Use when users want to wrap a session into a reusable component, extract workflows, make completed work reusable, abstract session work into patterns, or create a skill from session history.

practical-stack/ai-lab

Skill Claude CodeCodex

Guide for creating and validating AI agent skills based on Anthropic's official skill specification. Use when users want to create a new skill, update an existing skill, validate skills, run a skill audit, or need guidance on skill architecture and best practices. Do NOT use for creating agents (use meta-agent) or…

practical-stack/ai-lab

Instructions file CodexOpenCode

AGENTS.md instructions for practical-stack/ai-lab, covering ai lab: llm research workspace, purpose, directory structure, research workflow and phase 1: research.

practical-stack/ai-lab

Agent

Cost: EXPENSIVE Mode: Primary (orchestrator) Model: anthropic/claude-opus-4-5 Source: src/agents/atlas.ts.

practical-stack/ai-lab

Agent

Cost: FREE Mode: Read-only (subagent) Model: opencode/gpt-5-nano Source: src/agents/explore.ts.

practical-stack/ai-lab

Agent

Cost: CHEAP Mode: Read-only (subagent) Model: opencode/big-pickle Source: src/agents/librarian.ts.

practical-stack/ai-lab

Agent

Cost: EXPENSIVE Mode: Read-only (subagent) Model: anthropic/claude-sonnet-4-5 Source: src/agents/metis.ts.

practical-stack/ai-lab

Agent

Cost: EXPENSIVE Mode: Read-only (subagent) Model: anthropic/claude-sonnet-4-5 Source: src/agents/momus.ts.

practical-stack/ai-lab

Agent

Cost: CHEAP Mode: Read-only (subagent) Model: Vision-capable model Source: src/agents/multimodal-looker.ts.

practical-stack/ai-lab

Agent

Cost: EXPENSIVE Mode: Read-only (subagent) Model: openai/gpt-5.2 Source: src/agents/oracle.ts.

practical-stack/ai-lab

Agent

Cost: EXPENSIVE Mode: Subagent (planner) Model: anthropic/claude-opus-4-5 Source: src/agents/prometheus-prompt.ts Added: v3.1+.

practical-stack/ai-lab

Agent

Cost: EXPENSIVE Mode: Primary (orchestrator) Model: anthropic/claude-opus-4-5 Source: src/agents/sisyphus.ts Added: v3.1+.

practical-stack/ai-lab

Agent

Model: Opus (HIGH) File: agents/architect.md (148 lines) Tool Restrictions: disallowedTools: Write, Edit.