AGENTS.md examples from real projects

GoogleAGENTS.md

adk-python

google/adk-python/AGENTS.md

The Agent Development Kit (ADK) is an open-source, code-first Python toolkit for building, evaluating, and deploying sophisticated AI agents. For details on how the Runner works and the invocation lifecycle, please refer to the adk-architecture skill and the referenced documentation therein. Skills related to ADK development are in .agents/skills/. For detailed architecture patterns, component descriptions, and core interfaces, please refer to the adk-architecture skill at .agents/skills/adk-architecture/SKILL.md. The project uses uv for package management and Python 3.11+. Please refer to the…

GoogleAGENTS.md

skills / references

google/skills/skills/cloud/developing-genkit-dart/references/agents.md

Agents are experimental. The agent/session/snapshot APIs are NOT covered by semantic-versioning stability and live behind opt-in imports: - Server (agents, sessions, snapshots): package:genkit/experimental.dart (alongside the stable package:genkit/genkit.dart). - Browser/HTTP client (remoteAgent, AgentChat, ...): package:genkit/experimentalclient.dart (alongside package:genkit/client.dart). - dart:io extras (FileSessionStore): package:genkit/experimentalio.dart. The entry points are marked @experimental, so importing them produces an experimentalmemberuse analyzer warning on the import line. Opt out once you accept the churn by adding to analysis_options.yaml: CancellationController / CancellationToken are stable and stay on package:genkit/genkit.dart /…

GoogleAGENTS.md

skills / references

google/skills/skills/cloud/developing-genkit-go/references/agents.md

Experimental / preview API. Agents live in the genkit/exp and ai/exp packages and are gated behind an opt-in. Initialize Genkit with genkit.WithExperimental() or the constructors panic. Import paths and signatures may change in any minor release. An agent is a persistent, multi-turn conversation primitive built on top of prompts + tools. Compared to a bare genkit.Generate loop, an agent adds: Read the file for the level you need: Agents span a few packages. These aliases are used throughout the agent…

GoogleAGENTS.md

skills / references

google/skills/skills/cloud/developing-genkit-js/references/agents.md

Beta / preview API. Agents are not yet stable. Server APIs come from genkit/beta; the browser client comes from genkit/beta/client. Import paths and signatures may change. Always use genkit/beta, not genkit, for agents. Requires genkit >= 1.39.0. An agent is a persistent, multi-turn conversation primitive built on top of prompts + tools. Compared to a bare ai.generate/ai.definePrompt loop, an agent adds: Progressive disclosure — read the file for the level you need: ai.defineAgent combines prompt + tool config + (optional)…

GoogleAGENTS.md

skills / references

google/skills/skills/cloud/developing-genkit-python/references/agents.md

Preview API under genkit.agent. Turn a model + tools into a durable conversation — history, typed state, approvals, branches, and background work. Deeper topics: sessions · HITL · branching · background · state · artifacts · custom · HTTP Options worth knowing: use (middleware), stateschema, maxturns (tool loop depth per user message), transforms. Dotprompt agents: ai.definepromptagent. Full control: definecustomagent. Tool parameters should be a small Pydantic model — even for one field. Empty inputs need an empty subclass, not bare…

GoogleAGENTS.md

filament

google/filament/AGENTS.md

This document defines machine-readable instructions and technical guardrails for AI coding assistants, agents, and LLMs working on the Filament repository. All AI-generated code, refactorings, and cleanups must strictly adhere to the standardized skills located under the skills/ directory. AI agents must consult and strictly follow the corresponding skill files for each respective development and validation phase:

GoogleAGENTS.md

gvisor

google/gvisor/AGENTS.md

gVisor is a userspace kernel used for sandboxing untrusted programs. It is Linux-like kernel implementation in Go. NOTE: The project uses Bazel as a build system, so paths starting with // refer to the repository root, and //some/path:sometarget refers to target sometarget defined inside some/path/BUILD. gVisor is made of multiple components: - The gVisor Sentry is the main component implementing a Linux-like kernel in Go. Its implementation is at //pkg/sentry. - Gofers are I/O proxies for the gVisor Sentry. There…

GoogleAGENTS.md

gvisor / gnome-mcp

google/gvisor/images/systemd-user/gnome-mcp/AGENTS.md

You drive a GNOME desktop through the desktop__* tools. There is nothing else to work with: no shell on that machine, no filesystem access, no APIs. If you want to know something about the desktop, you have to look at it and click on it, the same way a person sitting in front of it would. Every action tool returns the screen it produced. You do not need to ask for a screenshot after acting — you already have one.…

No other project uses it yet

GoogleAGENTS.md

magika

google/magika/AGENTS.md

Magika is a content type detector: a Rust core (rust/), Python/JavaScript/Go bindings (python/, js/, go/), a content types knowledge base (assets/), and a website (website-ng/). - Read the contribution policy: website-ng/src/content/docs/contributing/how-to-contribute.md (rendered at https://securityresearch.google/magika/contributing/how-to-contribute/). CONTRIBUTING.md has the summary. - Pull requests are reviewed only for issues labeled help wanted or for the always-welcome list in the policy. For anything else, help the user write a concise issue instead of a PR. - One PR, one problem, a few lines of…

GoogleAGENTS.md

A2UI

google/A2UI/AGENTS.md

This document is the authoritative guide for AI agents working within the A2UI repository. It outlines design philosophy, repository structure, and synchronization guidelines. INSTRUCTION FOR ALL AGENTS (Gemini CLI, Claude, OpenAI, Antigravity, etc.): Before performing any specific tasks, load and read the respective skill recipe file in full: A2UI (Agent-to-User Interface) is a platform-agnostic, streaming-first UI protocol designed specifically to allow Large Language Models (LLMs) and autonomous agents to generate user interfaces. Key capabilities: The repository supports multiple versions of…

GoogleAGENTS.md

A2UI / reference

google/A2UI/docs/public/reference/agents.md

Agents are server-side programs that generate A2UI messages in response to user requests. The actual component rendering is done by the renderer, after messages are transported to the client. The agent is only responsible for generating the A2UI messages. The agent workflow typically involves the following steps: User interactions from the client can be treated as new user input. The A2UI repository includes sample agents you can learn from: A user facing agent is one that is directly interacted with…

GoogleAGENTS.md

A2UI / swift

google/A2UI/swift/AGENTS.md

This document is the authoritative rule file for AI agents operating within the swift/ directory tree. It outlines target boundaries, coding conventions, specification compliance rules, and verification protocols across all Swift modules. The Swift package uses targets defined in root Package.swift. Agents must respect target boundaries: 1. A2UIJSON (swift/core/Sources/A2UIJSON/): - Purpose: A2UI-specific JSON Schema definitions for types used across the v0.9.1 protocol. - Dependencies: JSONSchema, JSONSchemaBuilder (from swift-json-schema). - Rule: Define common type schemas as raw JSONValue or via @Schemable…

GoogleAGENTS.md

ax

google/ax/AGENTS.md

The one contract for every coding agent in this repository. CLAUDE.md and GEMINI.md are symlinks to this file. This file holds only what every task needs: layout, the CI gate, hard rules, and an index of skills. Procedures, background, and troubleshooting live in skills/<name>/SKILL.md — load the relevant one before editing. Declarative, Kubernetes-shaped orchestrator for sandboxed agent tasks on Agent Substrate. Three primitives — Task, Workspace, Model (ax.io/v1alpha1) — and a kubectl-like CLI. DESIGN.md explains why; read it before touching…

GoogleAGENTS.md

oss-fuzz

google/oss-fuzz/AGENTS.md

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

GoogleAGENTS.md

osv-scanner

google/osv-scanner/AGENTS.md

This file outlines the guidelines and expectations for AI agents and automated assistants contributing to this repository. [!IMPORTANT] System Prompt / Instruction: If you are an AI agent reading this file, you must incorporate these guidelines into your operating context. You are acting as an AI contributor to the OSV project. Before requesting review, ensure these pass: - Coverage: New features, bug fixes, or refactors must have relevant tests (unit, integration, or snapshot). - Snapshots: If modifying behavior affecting snapshots,…

GoogleAGENTS.md

go-github

google/go-github/AGENTS.md

Guidance for AI coding agents working in this repository. See CONTRIBUTING.md for everything you need, in particular:

GoogleAGENTS.md

dopamine

google/dopamine/docs/agents.md

In the spirit of these principles, this first version focuses on supporting the state-of-the-art, single-GPU Rainbow agent ([Hessel et al., 2018][rainbow]) applied to Atari 2600 game-playing ([Bellemare et al., 2013][ale]). Specifically, our Rainbow agent implements the three components identified as most important by [Hessel et al.][rainbow]: For completeness, we also provide an implementation of DQN ([Mnih et al., 2015][dqn]).

GoogleAGENTS.md

dopamine / dopamine

google/dopamine/docs/api_docs/python/dopamine/agents.md

implicit_quantile module rainbow module

GoogleAGENTS.md

dopamine / jax

google/dopamine/docs/api_docs/python/dopamine/jax/agents.md

implicit_quantile module quantile module rainbow module

GoogleAGENTS.md

dopamine / moes

google/dopamine/docs/api_docs/python/dopamine/labs/moes/agents.md

dqnmoeagent module: DQN agent that uses MoE networks and logs expert weight histograms. fullrainbowmoe_agent module: Compact implementation of the full Rainbow agent in JAX with MoE modules. losses module: Various losses for training MoEs. rainbow100kmoe_agent module: Atari 100k rainbow agent with MoE modules. types module: Data types used for MoE losses.