obraAGENTS.mdsuperpowers
obra/superpowers/AGENTS.md
theoretical fixes Every PR must solve a real problem that someone actually experienced. "My review agent flagged this" or "this could theoretically cause issues" is not a problem statement
AGENTS.md examples from real projects
How real projects brief Codex, Cursor and every other agent that reads AGENTS.md.
Best matchesWorked for most · soon
obraAGENTS.md
NousResearchAGENTS.mdhermes-agent / agent
NousResearch/hermes-agent/agent/AGENTS.md
name == ...` chain); `tools/todo_tool.py` is the pattern. ## Message-flow invariants (every change is reviewed against these) - **Prompt caching must not break.** Never alter past context, change toolsets, reload memories, or rebuild
affaan-mAGENTS.mdECC
affaan-m/ECC/AGENTS.md
design and scalability | Architectural decisions | | tdd-guide | Test-driven development | New features, bug fixes | | code-reviewer | Code quality and maintainability | After writing/modifying code | | security-reviewer | Vulnerability detection | Before commits, sensitive
DeepSeekAGENTS.mddeepseek-harness
deepseek-ai/deepseek-harness/docs/AGENTS.md
timing, ownership, modality, exceptions, consequences, and non-obvious orientation; delete narration, test walkthroughs, review analysis, and code restatement. Keep the local contract and link its rationale. Use [dsh-prose-standard
DeepSeekAGENTS.mddeepseek-harness / system
deepseek-ai/deepseek-harness/native/system/AGENTS.md
packages have one independent version and release workflow. ## Runtime rules - Landlock's argv, exit codes, diagnostics, and fail-closed confinement are defined in [docs/cli-contract.md](docs/cli-contract.md). Do not change them when
MicrosoftAGENTS.mdvscode / references
microsoft/vscode/extensions/copilot/assets/prompts/skills/agent-customization/references/agents.md
Supported Events `SessionStart`, `UserPromptSubmit`, `PreToolUse`, `PostToolUse`, `PreCompact`, `SubagentStart`, `SubagentStop`, `Stop` ### Example ```yaml --- description: "Secure code reviewer that blocks dangerous commands" tools: [read, search, execute] hooks: PreToolUse: - type: command command: "./scripts/block-dangerous-cmds.sh
MicrosoftAGENTS.mdvscode / copilotcli
microsoft/vscode/extensions/copilot/src/extension/chatSessions/copilotcli/AGENTS.md
Copilot CLI agent: * File Operations * Shell Operations * Search Operations * Agent & Task Operations * User Interaction * Code Review & Git * Data, Memory & MCP * Security ### `common/copilotCLIPrompt.ts` Parses raw user prompts and extracts structured chat
MicrosoftAGENTS.mdvscode / prompts
microsoft/vscode/src/vs/platform/agentHost/node/copilot/prompts/AGENTS.md
other prompt or filesystem inputs to this setting without an explicit security review. Code-managed prompt experiments should still use a contributor (Lever 2) gated
Hugging FaceAGENTS.mdtransformers
huggingface/transformers/.ai/AGENTS.md
Commands - `make style` — ruff format + lint (required for code style checks). - `make typing` — ty type checker and model structure rules. - `make fix-repo` — the `make style` fixes plus copies, modular
langgeniusAGENTS.mddify / web
langgenius/dify/web/AGENTS.md
load it for test-only, copy-only, or styling-only changes. - Use `frontend-code-review` only for explicit frontend review or audit requests, including test reviews. Use `frontend-testing` when
VercelAGENTS.mdnext.js / test
vercel/next.js/test/AGENTS.md
Test review instructions - During code review, flag any change that newly disables, skips, or omits deploy-mode tests or assertions, including `skipDeployment: true` and `isNextDeploy` conditionals. Only allow this when
GitHubAGENTS.mdspec-kit
github/spec-kit/AGENTS.md
covers registration, validation, execution, resume, and installed step packages. ## Testing Executable Behavior Before changing code or configuration that runs or controls execution without an LLM, read [Testing deterministic behavior](CONTRIBUTING.md
MicrosoftAGENTS.mdPowerToys
microsoft/PowerToys/AGENTS.md
Top-level AI contributor guidance for developing PowerToys - a collection of Windows productivity utilities
OpenAIAGENTS.mdcodex
openai/codex/AGENTS.md
your new functionality. Refactor existing code as necessary to make this happen. Likewise, when reviewing code, do not hesitate to push back on PRs that would unnecessarily add code
SupabaseAGENTS.mdsupabase
supabase/supabase/AGENTS.md
rules they apply) - `telemetry-standards` — PostHog events, `packages/common/telemetry-constants.ts` - `dev-toolbar-review` — `packages/dev-tools`, `packages/common/posthog-client.ts`, `packages/common/feature-flags.tsx` - `safe-sql-execution` — any code that builds or executes SQL against user databases - `react-hook-form
SupabaseAGENTS.mdsupabase / apps
supabase/supabase/apps/docs/AGENTS.md
style-guide/01-voice-and-tone.md` — person, tense, sentence length, brevity - `style-guide/02-elements.md` — admonitions, code blocks, procedures, tabs, images - `style-guide/03-page-structure.md` — document type, section grouping, chunking Load the file you need
SupabaseAGENTS.mdsupabase / studio
supabase/supabase/apps/studio/AGENTS.md
number inputs | `react-hook-form` | | Displaying API errors | `studio-error-handling` | | Tests (deciding, writing, reviewing) | `studio-testing`, then `studio-mock-api-tests` (component/MSW) or `studio-e2e-tests` (Playwright) | | PostHog event
MicrosoftAGENTS.mdterminal
microsoft/terminal/AGENTS.md
establish beyond reasonable doubt that a maintainer has permitted it. - Bug reports and code reviews, *IF* AI-use is disclosed *AND* a human has personally verified all findings and evidence
About AGENTS.md
What is AGENTS.md?
An open format for instructions to coding agents, read by Codex, Cursor and others. Think of it as a README written for agents.
Where does it go?
At the repository root, with more specific files in subdirectories. Agents read the one closest to the file they're editing.
What should it contain?
Setup and test commands, code style, and the rules a new contributor would need to know.
Does Claude Code read it?
Claude Code reads CLAUDE.md. A one-line CLAUDE.md that points at AGENTS.md covers both.