AGENTS.md examples from real projects

MicrosoftAGENTS.md

agentrc

microsoft/agentrc/AGENTS.md

@microsoft/agentrc is a CLI tool (and VS Code extension) that sets up repositories for AI-assisted development. It analyzes repos, generates instruction files, evaluates AI responses, and runs readiness reports. @agentrc/core (in packages/core/) is bundled into the CLI output — it is not an external npm package. Use the @agentrc/core/* path alias for all cross-package imports. Output is ESM ("type": "module"). All node_modules are external except @agentrc/core, @github/copilot-sdk, and vscode-jsonrpc (bundled via noExternal in tsup.config.ts). SDK shim: tsup.config.ts patches getBundledCliPath() in…

MicrosoftAGENTS.md

agentrc / well-tooled-rust

microsoft/agentrc/src/services/__tests__/fixtures/rust-policy/well-tooled-rust/AGENTS.md

Keep the workspace lint, formatting, test, and supply-chain evidence intact.

SentryAGENTS.md

sentry-go

getsentry/sentry-go/AGENTS.md

Single-module Go SDK with integration sub-modules in github.com/getsentry/sentry-go. AI commits MUST include: The root package sentry contains the entire public API. Type-safe key-value builders used by structured logging and metrics: Each lives in its own directory with a separate go.mod: When adding a new integration, mirror an existing one. Test tier preference (use the highest tier that covers what you need): Conventions: What to test: Thread safety:

MongoDBAGENTS.md

mongo-cxx-driver

mongodb/mongo-cxx-driver/AGENTS.md

This project uses CMake. The MongoDB C Driver (which provides the required bson and mongoc libraries) is fetched automatically via FetchContent unless an existing installation prefix is found via find_package() (set CMAKE_PREFIX_PATH accordingly). Use build/ as the CMake binary directory unless otherwise specified by the user. When the user specifies a custom binary directory, always use that directory - do not fall back to build/. The typical configure and build steps (for release and installation): The optional install step: [!IMPORTANT]…

NVIDIAAGENTS.md

cuopt

NVIDIA/cuopt/AGENTS.md

AI agent skills for NVIDIA cuOpt optimization engine. Skills live in skills/ (repo root) and use a flat layout: per domain, a concept skill (formulation / problem types) plus implementation skills — typically one per interface (Python, C, CLI, server), or consolidated when the content is shared across interfaces (e.g. installation). Skills evolve through agent interactions — see skills/cuopt-skill-evolution/ for the evolution workflow. 🔒 MANDATORY — Ambiguity: When the problem could be read more than one way, you MUST either…

No other project uses it yet

MicrosoftAGENTS.md

shell-intune-samples

microsoft/shell-intune-samples/AGENTS.md

Canonical entry point for AI agents and human contributors. Read this first. This repository is a curated collection of sample shell scripts maintained by the Microsoft Intune Customer Experience Engineering (CXE) team. The samples show how to manage macOS and Linux devices with Microsoft Intune. They are provided for education and "the art of the possible" — download, test, and adapt them, and run them only in a non-production / test environment. Classification: non-production (samples / learning environment). The full…

MicrosoftAGENTS.md

testfx

microsoft/testfx/AGENTS.md

Excluded-from-upstream skills and helpers are repository-owned and are NOT checked by gh copilot-curate verify; compare them explicitly on refresh. Details: .agents/skills/CURATION.md Use msbuild as the MSBuild entry point; nine legacy SKILL.md files are excluded from upstream curation. The grading rubric is not forked. Use curation-exceptions for gh copilot-curate update/verify and repository-owned skill edits. Run gh copilot-curate list to see installed plugins; run gh copilot-curate update to refresh.

MicrosoftAGENTS.md

testfx / common

microsoft/testfx/eng/common/AGENTS.md

Files under eng/common come from Arcade. Edits in eng/common will be overwritten by automation unless the changes are made directly in the Arcade repository. For more information, see the Arcade documentation.

MongoDBAGENTS.md

mongo-tools

mongodb/mongo-tools/AGENTS.md

We prefer that AI agents start by planning their work, documenting their plan as a Markdown file, and then iterating on that plan with the human operator. Don't just jump…

MicrosoftAGENTS.md

work-iq

microsoft/work-iq/AGENTS.md

Work IQ is a Copilot CLI plugin marketplace for managing AI agent plugins for GitHub Copilot CLI. It provides MCP servers, skills, and tools that connect AI assistants to Microsoft…

AppleAGENTS.md

servicetalk

apple/servicetalk/AGENTS.md

Guidance for AI coding agents working in the ServiceTalk repository. Humans should start with README.adoc and CONTRIBUTING.adoc. ServiceTalk is a JVM networking framework built on Netty, providing HTTP/1.1, HTTP/2, and gRPC clients and servers over an asynchronous Publisher/Single/ Completable core, with blocking and streaming programming models layered on top. It is a Gradle multi-project build of ~90 modules named servicetalk-<area>[-<impl>]: - servicetalk-<area>-api — public interfaces and types (e.g. http-api). - servicetalk-<area>-netty — the Netty-backed implementation (e.g. http-netty). - *-internal —…

AppleAGENTS.md

servicetalk / servicetalk-grpc-netty

apple/servicetalk/servicetalk-grpc-netty/AGENTS.md

Most tests in this module bind a real socket and run a real gRPC client against a real server, often over the HTTP/2 transport or a Unix domain socket. Before you add or change a test here, read ../.agents/skills/servicetalk-testing/references/transport-and-netty.md. It covers how to bind a server, how to share an ExecutionContext, and the order in which to close resources. GrpcUdsTest is the exemplar for this module. Start from ../.agents/skills/servicetalk-testing/SKILL.md for the conventions that apply to every test, and from the…

AppleAGENTS.md

servicetalk / servicetalk-http-netty

apple/servicetalk/servicetalk-http-netty/AGENTS.md

Most tests in this module bind a real socket and run a real HTTP client against a real server. They are the largest group of such tests in the repository. Before you add or change a test here, read ../.agents/skills/servicetalk-testing/references/transport-and-netty.md. It covers how to bind a server, how to share an ExecutionContext, the order in which to close resources, and when to use EmbeddedChannel instead. Start from ../.agents/skills/servicetalk-testing/SKILL.md for the conventions that apply to every test, and from the repository…

AppleAGENTS.md

servicetalk / servicetalk-transport-netty-internal

apple/servicetalk/servicetalk-transport-netty-internal/AGENTS.md

This module owns the shared transport test fixtures that the rest of the repository depends on: AddressUtils, ExecutionContextExtension, EmbeddedDuplexChannel, CloseUtils, MockFlushStrategy, and RandomDataUtils, all under src/testFixtures/java. Two consequences: - Its own tests mostly drive an EmbeddedChannel rather than a real socket. Prefer that when it can answer the question. - A change to anything in src/testFixtures affects many other modules. Check the callers before you change a signature. ../.agents/skills/servicetalk-testing/references/transport-and-netty.md documents these fixtures and how consumers use them. Keep it in…

SupabaseAGENTS.md

supabase-flutter

supabase/supabase-flutter/AGENTS.md

This file provides guidance to the agents when working with code in this repository. This is a monorepo for the Supabase Flutter SDK, managed by Melos. It contains multiple independently versioned packages that provide Flutter and Dart clients for Supabase services. The repository follows a layered dependency structure: - **supabase_flutter**: Flutter-specific wrapper with platform integrations (deep links, local storage, app lifecycle) - supabase: Core Dart client that orchestrates all service clients - **supabase_auth**: Authentication client (sessions, JWT, OAuth) - postgrest:…

No other project uses it yet

SentryAGENTS.md

sentry-cli

getsentry/sentry-cli/AGENTS.md

This is sentry-cli, a command-line utility for working with Sentry. It's primarily written in Rust with JavaScript/Node.js wrapper components for npm distribution. MUST follow Sentry's commit message format: type(scope): subject

SentryAGENTS.md

sentry-cli / workflows

getsentry/sentry-cli/.github/workflows/AGENTS.md

ALWAYS pin GitHub Actions to specific commit SHAs instead of version tags for enhanced security. This prevents supply chain attacks where tags could be moved to malicious commits. The comment…

SentryAGENTS.md

sentry-cli / lib

getsentry/sentry-cli/lib/AGENTS.md

No summary in the file. Open it to read it.

SentryAGENTS.md

sentry-cli / src

getsentry/sentry-cli/src/AGENTS.md

Always run cargo fmt --all, prior to committing any Rust code, to ensure consistent formatting and to avoid CI failures. Some other commands you may find useful:

BlockAGENTS.md

berd

block/berd/AGENTS.md

Guidelines for agents working on Berd. Berd is a standalone Tauri 2 + React 19 desktop. ACP is the main interface we use for the actual agent loop - creating and running sessions, finding available models, and setting configuration. When available, we work over ACP methods, but the UI can handle operations that are not yet in ACP or are client specific. - src/ — React UI/features/shared code - src-tauri/ — Tauri shell that starts or resolves goose serve -…