Guide new users through TAO framework setup, concepts, and first execution. Use when someone asks about TAO, how to get started, or needs help understanding the workflow.
Use when someone is new to TAO, asks how it works, or needs help setting up their first project.
TAO (Trace · Align · Operate) is an AI-native development framework that organizes coding with AI agents into three layers:
| File | Purpose |
|---|---|
CLAUDE.md | Project identity + code patterns (root) |
.github/tao/tao.config.json | Single source of truth for config |
.github/tao/CONTEXT.md |
| Current state — active phase, decisions |
.github/tao/CHANGELOG.md | History — what changed, when, by whom |
.github/tao/RULES.md | Framework rules (R0-R7) |
.github/skills/INDEX.md | Skill catalog (if exists) |
.github/agents/*.agent.md | Agent definitions |
.github/tao/tao.config.json — customize models, paths, lintCLAUDE.md — add project-specific rules and code patterns.github/tao/CONTEXT.mdchat.useCustomAgentHooks in VS Code settings@Execute-Tao and say "execute"@Execute-Tao → reads CONTEXT.md → finds active phase
→ reads PLAN.md → picks next task
→ routes to right model (Sonnet/Opus/GPT-4.1)
→ executes task → runs lint → updates CONTEXT.md
→ commits → picks next task → repeats
| Agent | When to use |
|---|---|
@Execute-Tao | Full autonomous execution loop |
@Brainstorm-Wu | Planning, ideation, trade-off analysis |
@Investigate-Shen | Complex debugging, architecture decisions, security audits |