How real projects brief Claude Code: the commands, conventions and warnings they put in CLAUDE.md.
codewithmukeshCLAUDE.mdcodewithmukesh/dotnet-claude-kit/CLAUDE.md
place. Respect token budgets. - **Every recommendation has a "why"** — No bare rules without justification - **Code examples must be modern C#** — Primary constructors, collection expressions, file-scoped namespaces, records - **No Swashbuckle
JordanCoinCLAUDE.mdJordanCoin/codemap/CLAUDE.md
moving code - Tracing imports or dependencies **ALWAYS run `codemap --diff` when:** - Reviewing or summarizing changes - Before committing code - User asks what changed - Use `--ref ` when comparing against something other than
qf-studioCLAUDE.mdqf-studio/pilot/CLAUDE.md
design - ✅ Create GitHub issues with `pilot` label for execution - ✅ Review every PR before merging - ❌ In *interactive* sessions, do not write code directly — defer to Pilot so the knowledge graph
Hugging FaceCLAUDE.mdhuggingface/Repo2RLEnv/CLAUDE.md
bootstrap sandbox | stable | 0002 | | `commit_runtime` | same, at commit level (bypasses PR-review filters) | stable | 0003 | | `code_instruct` | LLM authors a problem + verifier anchored to real repo source | experimental
partme-aiCLAUDE.mdpartme-ai/full-stack-skills/CLAUDE.md
guides, role definitions, ecosystem mappings - **Quality gate**: CI workflows that review SKILL.md changes on PRs There is no application code, build system, or test suite in this repo
ruvnetCLAUDE.mdruvnet/metaharness/kimi-k3-harness/CLAUDE.md
opus | Designs the change before code is written. | | `implementer` | sonnet | Writes code that matches the surrounding style. | | `reviewer` | opus | Hunts correctness bugs in the diff. | | `test-writer` | sonnet | Adds
JuliusBrusseeCLAUDE.mdJuliusBrussee/cavemem/CLAUDE.md
login`. - Conventional Commits for commit messages. PRs require passing CI and one review. ## Authorship voice Code comments are minimal and explain **why**, not what. Keep naming explicit. Prefer pure functions
mhattingpeteCLAUDE.mdmhattingpete/claude-skills-marketplace/CLAUDE.md
fixing, git-pushing, review-implementing - plan-implementer agent 2. **visual-documentation-skills** (v1.0.0) - visual-html-creator for stunning HTML documentation 3. **productivity-skills** (v1.0.0) - conversation-analyzer, code-auditor, project-bootstrapper
joaoh82CLAUDE.mdjoaoh82/rustunnel/CLAUDE.md
configured with an `http://` URL, e.g. Twilio) and 308-redirects everything else; `"redirect"` (the code default) 308s every request. All three edges run `"proxy"` — the live per-region files
jaredrhodCLAUDE.mdjaredrhod/ai-memory-vault/templates/CLAUDE.md
code edit.** Treat project source code as read-only by default. Before editing any code file, any config that affects a running system, or any commit / push / deploy, state
xjtulycCLAUDE.mdxjtulyc/MedgeClaw/CLAUDE.md
contained (base64-embed all images), professional CSS ## Scientific Skills 参考（K-Dense） **无论是否使用 Claude Code，遇到生物医药或科研场景任务时，必须主动查阅相关的 K-Dense Scientific Skills 作为参考。** Skills 位于 `scientific-skills/scientific-skills/` 目录，每个子目录包含一个 `SKILL.md
samvallad33CLAUDE.mdsamvallad33/vestige/CLAUDE.md
locally by default - MCP-native integration with coding agents - retrieval and memory lifecycle behavior informed by cognitive science - explicit tools for search, review, suppression, purge, graph exploration, contradiction inspection
umputunCLAUDE.mdumputun/agterm/CLAUDE.md
parity. - Cookbook recipes are third-party work published by their author, not code the project owns. Review asks only three things: it does no harm, deliberately or accidentally; it does
tuannvmCLAUDE.mdtuannvm/codex-mcp-server/CLAUDE.md
session support, model selection, reasoning effort, sandbox mode, full-auto, and working directory options - **review**: Code review for uncommitted changes, branches, or commits - **websearch**: Web search using Codex CLI with
Matt-DionisCLAUDE.mdMatt-Dionis/claude-code-configs/claude-config-composer/CLAUDE.md
import what you use - Components are tree-shakeable - No runtime overhead from library ## 2. Code Splitting ```tsx // Lazy load heavy components const HeavyChart = lazy(() => import('@/components/ui/chart')) <Suspense fallback={ }> ``` ## Form Controls
danielealbanoCLAUDE.mddanielealbano/android-remote-control-mcp/CLAUDE.md
implemented and all quality gates pass (linting, tests, build), you MUST ALWAYS spawn the `code-reviewer` subagent in plan compliance mode to verify the ENTIRE implementation matches the plan
emiliaprotocolCLAUDE.mdemiliaprotocol/emilia-protocol/CLAUDE.md
pull request for reviewable changes, and do not treat owner bypass as independent review. Until a second human maintainer exists, any owner-bypass direct push must be locally verified
rome-osCLAUDE.mdrome-os/rome/docs/CLAUDE.md
contract. The second is the mechanism that enforces it, and it belongs in code review. **Design context written into `docs/` outlives the change and rots there.** The design context
FerroxLabsCLAUDE.mdFerroxLabs/wayland/CLAUDE.md
think-build-ship end to end and dispatches peer specialist skills (frontend-design, TDD, code-review) as subagent tools when the workflow needs them
StuMasonCLAUDE.mdStuMason/coolify-mcp/CLAUDE.md
CLAUDE.md This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository. ## Project Overview MCP (Model Context Protocol) server for Coolify that provides 45 token
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.
