This folder contains context files optimized for AI coding assistants. Use the file that matches your preferred tool.
| Your Tool | File | To Activate |
|---|---|---|
| Claude Code | CLAUDE.md |
Copy to project root |
| Gemini CLI | GEMINI.md |
Copy to project root |
| Cursor | .cursor/rules/ |
Copy to project root |
| GitHub Copilot | .github/ |
Copy to project root |
| OpenAI Codex / Jules | AGENTS.md |
Copy to project root |
| Any LLM | llms.txt |
Use directly |
All context files are in this .ai/ folder. Copy the ones you need to your project root to activate them.
CLAUDE.md- Claude Code context with quick start, architecture, and common tasksGEMINI.md- Gemini CLI context with @import references to shared modulesAGENTS.md- Universal format for OpenAI Codex, Copilot, Cursor, Julesllms.txt- Minimal LLM-optimized summary for any AI assistant
architecture.md- Pipeline flow diagram and file structurecommands.md- All Makefile commands with descriptionsagents.md- Detailed reference for each agenttools.md- Custom tools documentationstate-keys.md- Session state key referencecommon-tasks.md- Step-by-step guides for adding agents, tools, callbackstroubleshooting.md- Common errors and fixes
The blog/ directory contains a 9-part progressive tutorial:
- Part 1: Setup and First Agent
- Part 2: IntakeAgent - Request parsing
- Part 3: MarketResearchAgent - Google Search
- Part 4: CompetitorMappingAgent - Google Maps API
- Part 5: GapAnalysisAgent - Code execution
- Part 6: StrategyAdvisorAgent - Extended reasoning
- Part 7: ArtifactGenerationPipeline - Multimodal outputs
- Part 8: Testing your agent
- Part 9: Production deployment
- Bonus: AG-UI interactive frontend
DEVELOPER_GUIDE.md- Deep architecture documentation- ADK Docs: https://google.github.io/adk-docs/
# Install dependencies
make install
# Run agent at localhost:8501
make dev
# Run tests
make test-unitThe .claude/skills/ folder contains skills that Claude automatically loads when relevant:
| Skill | Purpose | Auto-triggers When |
|---|---|---|
adk-agent-builder |
Add new agents | "add agent", "create agent" |
adk-tool-builder |
Create custom tools | "add tool", "new tool" |
adk-debugger |
Troubleshoot errors | "error", "not working", "debug" |
retail-agent-learner |
Learn ADK concepts | "explain", "how does", "architecture" |
retail-agent-customizer |
Customize for use cases | "customize", "adapt", "modify for" |
Skills vs Commands: Skills are model-invoked (Claude loads them automatically based on context). Commands are user-invoked (you type /command).
The .claude/commands/ folder contains slash commands for Claude Code:
/add-agent- Guide for adding a new agent/add-tool- Guide for adding a new tool/run-tests- Run the test suite/explain-pipeline- Explain the agent pipeline
To use: copy .ai/.claude/ to your project's root .claude/ folder, or symlink it.
All AI context files are stored in this .ai/ folder. To use them:
cp .ai/CLAUDE.md ./CLAUDE.md
cp -r .ai/.claude ./.claudeClaude Code automatically reads CLAUDE.md at project root.
cp .ai/GEMINI.md ./GEMINI.mdRun /memory add file:GEMINI.md to add to memory.
cp -r .ai/.cursor ./.cursorCursor reads .cursor/rules/*.mdc files automatically.
cp -r .ai/.github ./.githubCopilot reads .github/copilot-instructions.md automatically.
cp .ai/AGENTS.md ./AGENTS.mdFollows the universal AGENTS.md specification.
When updating agent architecture:
- Update
context/architecture.mdwith new pipeline flow - Update
context/agents.mdwith agent details - Update
context/state-keys.mdif new state keys added - Update main context files (CLAUDE.md, GEMINI.md, AGENTS.md)