How real projects brief Claude Code: the commands, conventions and warnings they put in CLAUDE.md.
Nano-CollectiveCLAUDE.mdNano-Collective/nanocoder/CLAUDE.md
start # Run the compiled application pnpm run dev # Watch mode compilation (tsc --watch) # Testing (run before committing) pnpm run test:all # Full suite: format, lint, types, AVA tests, knip, audit
VercelCLAUDE.mdvercel/ai-elements/.claude/CLAUDE.md
prop instead of `React.forwardRef` **Solid/Svelte/Vue/Qwik:** - Use `class` and `for` attributes (not `className` or `htmlFor`) --- ## Testing - Write assertions inside `it()` or `test()` blocks - Avoid done callbacks in async tests - use async/await
AgentsMeshCLAUDE.mdAgentsMesh/AgentsMesh/CLAUDE.md
docker | | Jaeger UI | localhost:10014 | docker | Each worktree adds offset×50 to every slot. Test accounts: - **User**: dev@agentsmesh.local / devpass123 - **Admin**: admin@agentsmesh.local / adminpass123 ### Logs ```bash tail -f deploy/dev/runtime/backend/backend.log # ibazel + backend stdout
redhat-etCLAUDE.mdredhat-et/ripwire/CLAUDE.md
shorter, and you already know it. Run gates in the **foreground**. A new `test/*check.sh` must be listed in `test/regression.sh` in the same commit — `test/manifestcheck.sh` fails otherwise. ## The five guardrails
nizosCLAUDE.mdnizos/tdd-guard/CLAUDE.md
Guard ## Project Goal TDD Guard is a Claude Code hook that enforces Test-Driven Development by intercepting file operations. When Claude Code attempts to edit or write files, TDD Guard
antonbabenkoCLAUDE.mdantonbabenko/terraform-skill/CLAUDE.md
owned, never hand-edit ├── mcp.json # Optional Kiro MCP (read-only terraform-mcp-server) ├── tests/ # Baseline scenarios and rationalization tracking │ ├── baseline-scenarios.md │ ├── compliance-verification.md │ └── rationalization-table.md └── .github/ ├── release/ │ ├── pre-commit.js # Release version-sync hook (SKILL.md
jnMetaCodeCLAUDE.mdjnMetaCode/agency-orchestrator/CLAUDE.md
prompt optimize " " # AI-optimize a prompt (--mode system|user, --save) ao prompt test / list / show / rm / garden # Prompt Lab: test / manage / starter templates ao skills [name] # List / view methodology skills
InfisicalCLAUDE.mdInfisical/agent-vault/CLAUDE.md
Typically deployed to a remote host that agents reach over the network. ## Build, run, test ```bash make build # Builds frontend (React/Vite) then Go binary → ./agent-vault make web-dev # Frontend-only
MetaCLAUDE.mdfacebookresearch/fairchem/CLAUDE.md
multi-task training via TorchTNT. ## Development Commands ```bash # Install pip install -e packages/fairchem-core[dev] # Tests (always pass -c flag) pytest tests -c packages/fairchem-core/pyproject.toml pytest tests/core/models/test_uma.py -vv pytest tests/core
agno-agiCLAUDE.mdagno-agi/dash/CLAUDE.md
Database URL builder evals/ # Evaluation framework (Agno eval classes) ├── run.py # Unified eval runner └── cases/ # Test cases by category ├── accuracy.py # AccuracyEval — data correctness ├── routing.py # ReliabilityEval — team routes correctly ├── security.py # AgentAsJudgeEval
muxy-appCLAUDE.mdmuxy-app/muxy/CLAUDE.md
Formatting Requires `swiftlint` and `swiftformat` (`brew install swiftlint swiftformat`). ```bash scripts/checks.sh # Format, lint, build, test scripts/checks.sh --fix # Auto-fix formatting and linting issues scripts/checks.sh --coverage # Also run the coverage gate
tractorjuiceCLAUDE.mdtractorjuice/arc-kit/CLAUDE.md
Build & Development Commands ```bash # Install in development mode (uv recommended) uv pip install -e . # Test CLI arckit --help arckit check # Test project initialization arckit init test-project --ai codex
MicrosoftCLAUDE.mdmicrosoft/kernel-memory/archived/v2/CLAUDE.md
warnings or errors - [ ] `coverage.sh` runs successfully without warnings or errors - [ ] there are zero skipped tests - [ ] `KNOWN-ISSUES.md` is up to date - [ ] if a PR exists, the PR title, description and content
HermannBjorgvinCLAUDE.mdHermannBjorgvin/Clawdmeter/CLAUDE.md
shared `chime.cpp` engine is ready to wire up once it's tested on hardware. - **No IO expander** despite the Waveshare wiki FAQ implying one. The schematic shows Key3/PWR wired directly
LangChainCLAUDE.mdlangchain-ai/agents-from-scratch/CLAUDE.md
notebooks/evaluation.ipynb` - Agent evaluation - `notebooks/hitl.ipynb` - Human-in-the-loop functionality - `notebooks/memory.ipynb` - Adding memory capabilities ## Running Tests ### Testing Scripts Test to ensure all implementations work: ```bash # Test all implementations python tests/run_all_tests.py
AmoyLabCLAUDE.mdAmoyLab/Unla/CLAUDE.md
Interface**: `web/` - React/TypeScript dashboard for configuration and chat ## Development Commands ### Go Backend ```bash # Run tests make test make test-coverage # with coverage report make test-race # with race detection # Build
iamzhihuixCLAUDE.mdiamzhihuix/skills-manage/CLAUDE.md
启动 Vite 开发服务器（端口 24200，单独前端调试用） pnpm build # TypeScript 编译 + Vite 构建 pnpm test # Vitest 运行全部测试（370+，含 3 个遗留的 Sidebar/Settings/PlatformIcon 失败，非本次改动引入） pnpm test -- src/test/skillStore.test.ts # 运行单个测试文件 pnpm test:watch
jhlee0409CLAUDE.mdjhlee0409/claude-code-history-viewer/CLAUDE.md
ESLint just tauri-build # Build production app (macOS universal binary, Linux native) just test # Run vitest in watch mode just test-run # Run tests once with verbose output just sync
mrgoonieCLAUDE.mdmrgoonie/claudekit-skills/CLAUDE.md
scripts requiring packages. - **Env var hierarchy**: process env > skill .env > skills/.env > .claude/.env > root .env ## Testing ```bash # Validate skill structure ~/.claude/skills/.venv/bin/python3 ~/.claude/skills/skill-creator/scripts/quick_validate.py .claude/skills/{name} # Run skill-specific tests cd .claude/skills/{name
louislvaCLAUDE.mdlouislva/claude-peers-mcp/CLAUDE.md
Use Bun instead of Node.js, npm, pnpm, or vite.
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.
