How real projects brief Claude Code: the commands, conventions and warnings they put in CLAUDE.md.
muthuishereCLAUDE.mdmuthuishere/mcp-server-bash-sdk/CLAUDE.md
failure, `-32022` unsupported version. `-32020..-32099` is reserved for the spec — do not invent codes there. Not implemented, deliberately: SSE response mode, resources, prompts, MRTR `InputRequiredResult`s, `subscriptions/listen` streams
aqm857886159CLAUDE.mdaqm857886159/Nomi/CLAUDE.md
自动调用只读、限时 `/ponytail-review` 适配器；超时按 diff 与负载派生、全机串行一把锁；失败或缺少结果 fail-closed，runner 不可用时只许 `PONYTAIL_REVIEW_DEFER=1` 留痕延后 | | R26 | 分层边界不许反向/循环 | 渲染层禁直捅主进程（走 bridge/中立契约层）、主进程禁反向 import 渲染层、禁新增完全静态循环；`check
1mcp-appCLAUDE.md1mcp-app/agent/CLAUDE.md
During refactoring, preserve validation, short-circuit order, side effects, and failure behavior. Review new or changed code against these rules and verify the affected behavior. ### Environment Variables - Access through yargs
bitwize-music-studioCLAUDE.mdbitwize-music-studio/claude-ai-music-skills/CLAUDE.md
command for the user's OS) - Skills stale: "⚠️ N skill(s) missing from Claude Code, N ghost — run: `claude plugin update bitwize-music`" - Album ideas (from `get_ideas`) - In-progress
mixpeekCLAUDE.mdmixpeek/amux/CLAUDE.md
create their own. Serialize task writes and retain task evidence independently. Publish the reviewed project result once, then expire its workers and remove only the published checkout. ## Isolated workers Isolated
ankitvguptaCLAUDE.mdankitvgupta/exo/CLAUDE.md
delegation failure as stopping at `gh pr create`. - **Re-run pre-PR periodically during code review** when there have been substantial changes since the last run. The first
trailofbitsCLAUDE.mdtrailofbits/skills-curated/CLAUDE.md
inject unrelated behavior - Obfuscated code (base64, eval, encoded strings) - Skills that are purely self-promotional with no real content - Duplicate submissions of existing skills ### Review Process - At least one CODEOWNER
jbarbierCLAUDE.mdjbarbier/CLAUDE.md/CLAUDE.md
task matches a specialized domain (SEO, schema, security audit, design review, etc.), use the installed Claude Code skill. Don't reinvent what gstack or a community skill already does well
bvisibleCLAUDE.mdbvisible/mcp-ssh-manager/CLAUDE.md
configuration - Manages `.env` file with server configurations - Tests connections using Paramiko - Configures Claude Code integration 3. **Deployment Helpers** (`src/deploy-helper.js`, `src/server-aliases.js`): Advanced features - Automated deployment strategies with permission handling - Server alias
claude-did-thisCLAUDE.mdclaude-did-this/claude-hub/CLAUDE.md
with the successful check suite - **Process**: Claude performs security, logic, performance, and code quality analysis - **Output**: Detailed review comments, line-specific feedback, and approval/change requests - **Integration**: Uses GitHub
davekilleenCLAUDE.mddavekilleen/Dex/CLAUDE.md
directions, and tasks you make in [app] come back through your daily plan for review. Takes [30 sec–2 min] to connect. Want to? (Run `/[app]-setup` anytime
davekilleenCLAUDE.mddavekilleen/Dex/.claude/plugins/compound-engineering/CLAUDE.md
README.md tables accurate (agents, commands, skills) - [ ] plugin.json description matches current counts ### Directory Structure ``` agents/ ├── review/ # Code review agents ├── research/ # Research and analysis agents ├── design/ # Design and UI agents ├── workflow/ # Workflow
HainrixzCLAUDE.mdHainrixz/the-architect/CLAUDE.md
version of it here. Standing rule, whatever the entry point: **never propose rewriting working code the user did not ask you to touch**, and the repo's existing conventions beat
IvanWng97CLAUDE.mdIvanWng97/pixtuoid/CLAUDE.md
just preflight` or one at a time. Never pipe preflight through `tail`/`head` (exit code eaten). - Touched `--json` / `SourceStatus` / `OutcomeRow` / the source roster → `just gen-contract` (regenerates schemas + Raycast types
aj-geddesCLAUDE.mdaj-geddes/claude-code-bmad-skills/CLAUDE.md
skill, agent, hook, or script here may write application code, run tests, lint, check coverage, build, or review an implemented diff. The last artifact the plugin produces is a `ready
fdmtlCLAUDE.mdfdmtl/director/CLAUDE.md
Read Before Writing**: Always read existing code to understand patterns before making changes - **Small Commits**: Make focused, atomic changes that are easy to review - **No Default Exports**: Enforced by Biome
HainrixzCLAUDE.mdHainrixz/whatsapp-agentkit/CLAUDE.md
Sistema de Instrucciones para Claude Code > Este archivo es el CEREBRO de AgentKit. Claude Code lo lee automáticamente > y sabe exactamente qué hacer para guiar al usuario a construir
ZaxbyHubCLAUDE.mdZaxbyHub/opencode-swarm/CLAUDE.md
Read `GRAPH_REPORT.md` (next to graph.json) only for broad architecture review, or when query/path/explain do not surface enough context. - After modifying code, run `graphify update .` **from the primary checkout** to refresh
rz1989sCLAUDE.mdrz1989s/claude-code-statusline/CLAUDE.md
CONFIG_PRAYER_LOCATION_MODE=local_gps ./statusline.sh ENV_CONFIG_LOCATION_FORMAT=full ./statusline.sh ``` ## Claude Code JSON Input Format (stdin) The statusline reads JSON from stdin (`input=$(cat)`), exported as `STATUSLINE
umputunCLAUDE.mdumputun/cc-thingz/CLAUDE.md
custom skills, actions, or configurations that exist only in a user's personal Claude Code setup. All documentation must refer only to what exists in this repository. ## Conventions - Hook scripts
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.
