How real projects brief Claude Code: the commands, conventions and warnings they put in CLAUDE.md.
AnthropicCLAUDE.mdanthropics/anthropic-sdk-typescript/CLAUDE.md
Bundlers follow every statically-resolvable import — including lazy import('./relative/path') — and fail browser builds on node: builtins in the resulting chunks.
AnthropicCLAUDE.mdanthropics/cwc-workshops/agent-battle/CLAUDE.md
You're helping a participant in a 45-minute hands-on workshop about Claude Managed Agents. They're competing to configure an agent (in my_agent.py) that mines the most diamonds in Minecraft. The point is for them to learn how CMA's levers work — system prompt, skills, MCP servers, model choice — not just to win. - When asked "how do I get more diamonds," explain the levers (what does SKILLS=[SKILLMINING] do? what's the trade-off vs writing your own prompt?) rather than handing over…
AnthropicCLAUDE.mdanthropics/cwc-workshops/agent-decomposition/.claude/CLAUDE.md
You're pairing with someone who's here to learn the tool → skill → subagent decision framework by decomposing an inventory agent. They'll ask you to make edits. Make them — but treat every edit as a teaching moment. 1. Explain before you edit. One short paragraph: what's wrong with the current state, what the change does, and which framework principle it maps to (compute-over-context / typed-contract / policy-as-skill / scope). 2. Show, don't just tell. Ground the lesson in something…
AnthropicCLAUDE.mdanthropics/cwc-workshops/eval-driven-agent-development/CLAUDE.md
When asked to setup the folder/repository, make sure the prerequisites listed in the README.md are met (Node >=22 installed, ant CLI installed, Docker installed, ANTHROPICAPIKEY set in environment). Do not deploy the resources yet. Offer to install them for the user if missing (if Docker is missing, point them to the installation website) Kick off docker build -t cwc-pptx-render . asap in a subagent to make sure the docker image is pre-built. Write the ANTHROPICAPIKEY from your environment into a…
AnthropicCLAUDE.mdanthropics/launch-your-agent/CLAUDE.md
No summary in the file. Open it to read it.
A file Claude Code reads at the start of every session. It holds the commands, conventions and warnings the agent needs for this project.
At the repository root. Claude Code also reads CLAUDE.md files in subdirectories when it works there.
Build and test commands, the project's layout, conventions that aren't obvious from the code, and mistakes to avoid. Short files tend to work better than long ones.
Claude Code reads CLAUDE.md; most other agents read AGENTS.md. Many projects keep one and point the other at it.
