AGENTS.md examples from real projects

ray-projectAGENTS.md

ray

ray-project/ray/AGENTS.md

welcome. ### Accountability - Pure code-agent PRs are **not allowed**. A human submitter must understand and defend the change end-to-end. - The submitting human must review every changed line

No other project uses it yet

ray-projectAGENTS.md

ray / doc

ray-project/ray/doc/AGENTS.md

doc/BUILD.bazel` and the example `BUILD.bazel` files emit it, because they define test targets. Library code tags select the same steps. - **`doc_example`** is what doc assets emit: `.py`, `.ipynb`, `.yaml

No other project uses it yet

HeyPuterAGENTS.md

puter

HeyPuter/puter/AGENTS.md

Understand what you commit.** AI assistance is fine; shipping code you couldn't have written, debugged, or defended in review is not. - **A limit change is not done until

gradio-appAGENTS.md

gradio

gradio-app/gradio/AGENTS.md

template. Trivial autocomplete does not need to be disclosed. All AI-generated code must be self-reviewed. 4. **Format your code before pushing.** - Backend: `bash scripts/format_backend.sh` - Frontend: `bash scripts/format_frontend.sh

No other project uses it yet

THU-MAICAGENTS.md

OpenMAIC

THU-MAIC/OpenMAIC/AGENTS.md

runs on every non-draft pull request. The contribution workflow is in [CONTRIBUTING.md](CONTRIBUTING.md). ## Code Review Rules ### Severity - P1 is only for problems that produce wrong behavior at runtime: user

No other project uses it yet

LizardByteAGENTS.md

Sunshine

LizardByte/Sunshine/AGENTS.md

This limit applies only to creating issues; investigate or fix issues fully when asked. ## Code Review Rules Check new and modified tests for the common-header and include-group conventions

sxyaziAGENTS.md

yazi

sxyazi/yazi/AGENTS.md

child crate is not a new workspace dependency. Evaluate new crates by the maintained code and correctness obligations they remove versus their integration, compatibility, security, and upkeep costs; add them

No other project uses it yet

deepspeedaiAGENTS.md

DeepSpeed

deepspeedai/DeepSpeed/AGENTS.md

enforced by the required CI check; do not report commit-signoff findings in code reviews. - Formatting: yapf (column_limit=119, `.style.yapf`) + flake8 (`.flake8`). - Always verify changed files pass pre-commit

No other project uses it yet

tinyhumansaiReads credentialsAGENTS.md

openhuman

tinyhumansai/openhuman/AGENTS.md

# OpenHuman OpenHuman is a React and Tauri v2 desktop assistant with an in-process

No other project uses it yet

wshobsonAGENTS.md

agents

wshobson/agents/docs/agents.md

completeness, consistency, and correctness | ### Quality Assurance & Security #### Code Quality & Review | Agent | Model | Description | | ------------------------------------------------------------------------------------------------ | ----- | --------------------------------------------------------------- | | [code-reviewer](../plugins/comprehensive-review/agents/code-reviewer.md) | opus | Code review with security focus and production reliability | | [security-auditor](../plugins/comprehensive-review/agents/security-auditor.md) | opus | Vulnerability

No other project uses it yet

fastlaneAGENTS.md

fastlane

fastlane/fastlane/AGENTS.md

comments, and do what it asks first (an audit, a discussion) before changing code. A review checks the change against it. - One concern per pull request. - A fix comes with

No other project uses it yet

qbittorrentAGENTS.md

qBittorrent

qbittorrent/qBittorrent/AGENTS.md

Wikipedia article]: https://en.wikipedia.org/wiki/BitTorrent ## Communication Use clear, appropriate English for non-native readers. ## Code Review Guidelines * Respect project configuration. \ Avoid reading excluded files unless it is strictly necessary

vimAGENTS.md

vim

vim/vim/AGENTS.md

runtime/syntax/testdir` - Tests for Indent runtime are in `runtime/indent/testdir` ## Common gotchas - **Distinguish what code enforces from what docs claim.** If a patch changes documented behavior, say so in the Problem/Solution. - **Generated

No other project uses it yet

ToolJetReads credentialsAGENTS.md

ToolJet

ToolJet/ToolJet/AGENTS.md

changes. Keep evidence links current, preserve explicit inference labels, and review map changes with the same owners as the code. Stale context is worse than no context. ## Agent skills ### Issue

No other project uses it yet

ToolJetAGENTS.md

ToolJet / frontend

ToolJet/ToolJet/frontend/AGENTS.md

that still passes against broken production code certifies nothing. **Full reference: `src/test/README.md`** (all frontend code). Review rules: `.github/instructions/frontend-tests.instructions.md`. - **Stack:** Jest + Testing Library + MSW. - **Layout:** - specs are `*.spec.*` in a colocated

No other project uses it yet

HmbownReads credentialsAGENTS.md

DeepSeek-TUI

Hmbown/DeepSeek-TUI/AGENTS.md

lane is obsolete, preserve its intent and evidence rather than merging stale code mechanically. - Agent-owned work lands on current `main`. Each writing agent works in its own short-lived

HmbownAGENTS.md

DeepSeek-TUI / web

Hmbown/DeepSeek-TUI/web/AGENTS.md

dictionary/content spine; do not add page-local language booleans, `isZh` copy forks, or hard-coded translated page clones. A compatibility exception must name the external contract that still requires

GitHubAGENTS.md

awesome-copilot

github/awesome-copilot/AGENTS.md

your contribution works with GitHub Copilot - [ ] Checked that file names follow the naming convention ### Code Review Checklist For instruction files (\*.instructions.md): - [ ] Has markdown front matter - [ ] Has non-empty `description` field

GitHubAGENTS.md

awesome-copilot / templates

github/awesome-copilot/skills/setup-my-iq/assets/templates/AGENTS.md

voice, formatting rules | `communicationStyle` | | Working preferences, hard rules | `preferencesAndConstraints` | | "Tell me about myself" / broad review | All six | Cross-cutting questions (e.g., "what do I do for team X?") may require

No other project uses it yet

HmbownReads credentialsAGENTS.md

CodeWhale

Hmbown/CodeWhale/AGENTS.md

lane is obsolete, preserve its intent and evidence rather than merging stale code mechanically. - Agent-owned work uses the canonical checkout on current `main`. Do not create branches, worktrees, nested