How real projects brief Codex, Cursor and every other agent that reads AGENTS.md.
MicrosoftAGENTS.mdmicrosoft/vscode/AGENTS.md
This file provides instructions for AI coding agents working with the VS Code codebase. For detailed project overview, architecture, coding guidelines, and validation steps, see the Copilot Instructions.
MicrosoftAGENTS.mdmicrosoft/vscode/extensions/copilot/assets/prompts/skills/agent-customization/references/agents.md
Custom personas with specific tools, instructions, and behaviors. Use for orchestrated workflows with role-based tool restrictions. Use reasoning-effort to set the reasoning level for the custom agent's model: The selected model must support the configured level. If omitted, the runtime resolves the effort from the model configuration and, when applicable, the parent agent. Sources: built-in aliases, specific tools, MCP servers (<server>/*), extension tools. Special: [] = no tools, omit = defaults. Body reference: #tool:<name> To discover available tools, check your…
MicrosoftAGENTS.mdmicrosoft/vscode/extensions/copilot/src/extension/chatSessions/copilotcli/AGENTS.md
This folder contains the Copilot CLI integration for VS Code Chat. It enables users to open a new Chat window and interact with a Copilot CLI agent instance directly within VS Code. VS Code provides the UI, Copilot CLI SDK provides the smarts. Important: The Copilot CLI agent functionality is powered by the @github/copilot/sdk package. See the SDK package for full type definitions. The integration follows VS Code's platform layering pattern with three layers:
MicrosoftAGENTS.mdmicrosoft/vscode/extensions/copilot/src/platform/authentication/common/AGENTS.md
IAuthenticationService manages GitHub and Copilot authentication. It provides GitHub sessions (OAuth tokens) and Copilot tokens (CAPI tokens). getGitHubSession requires a kind parameter. Choose thoughtfully: Returns AuthenticationSession (never undefined). Throws if the user cancels. Requires createIfNone with a StrictAuthenticationPresentationOptions containing a localized detail string explaining why auth is needed: Same as above but forces re-authentication even if a session exists. Use when the current token has lost authorization: Returns AuthenticationSession | undefined. Never shows UI. Use when auth is optional: For…
MicrosoftAGENTS.mdmicrosoft/vscode/src/vs/platform/agentHost/AGENTS.md
Node runtime service construction is documented separately in node/serviceBootstrapping.md. Host startup timings, process-lifetime correlation, and session workload counts are documented in PERFORMANCE.md. Status: COMPLETE (2026-07-01) All waves A–D and gates G-B1, G-C1, G-C2, G-D1 are done. Codex, Claude, and Copilot all use the unified orchestrator path. Codex advertises multipleChats: { fork: true, sideChat: true }. Model-backed peer, fork, and side-chat parity scenarios run in deterministic replay. The operational chat surface (send/abort/model/agent/history) is fully chat-addressed and uniform across harnesses. Session ownership…
MicrosoftAGENTS.mdmicrosoft/vscode/src/vs/platform/agentHost/common/state/AGENTS.md
This directory contains the VS Code-facing wrappers around the Agent Host Protocol (AHP) state model. Read this before modifying protocol types. The generated protocol includes the Changesets model: Optional protocol fields are usually backwards-compatible. Add the field in the source protocol repo, sync protocol/**, and add wrapper exports only when VS Code code needs them. Adding a new server-produced action type is backwards-compatible when old clients can ignore it safely. Old clients at the same protocol version ignore unknown action…
MicrosoftAGENTS.mdmicrosoft/vscode/src/vs/platform/agentHost/node/copilot/prompts/AGENTS.md
This directory customizes the system prompt for Copilot CLI agent host (ahp+cli) sessions. Read this before changing how the system message is built or adding per-model / per-tool guidance. It mirrors the Copilot extension's extensions/copilot/.../prompts/node/agent/ (agentPrompts), but the agent host runs in its own process and cannot use prompt-tsx, so contributors return plain data the SDK accepts directly. Unless the matching chat.agentHost.copilot.modelCapabilityOverrides entry provides a YAML promptOverrideString or promptOverrideFile containing systemPrompt, resolveSystemMessageConfig(model, context) layers, in order: The per-model prompt override…
MicrosoftAGENTS.mdmicrosoft/vscode/src/vs/platform/agentHost/test/node/AGENTS.md
For tests in this area that touch the SessionDatabase, they MUST use an in-memory database, not a real database file on disk. Use SessionDatabase.open(':memory:') and see the examples from existing tests.
MicrosoftAGENTS.mdmicrosoft/vscode/src/vs/workbench/contrib/imageCarousel/AGENTS.md
A generic workbench editor for viewing collections of images in a carousel/slideshow UI. Opens as a modal editor pane with navigation arrows, a caption, and a thumbnail strip. The image carousel is a self-contained workbench contribution that follows the custom editor pattern: Clicking an image attachment pill in chat (when chat.imageCarousel.enabled is true) executes the workbench.action.chat.openImageInCarousel command, which collects request attachment images together with response-derived images for the current chat session and opens them in the carousel. MIME types are…
Hugging FaceAGENTS.mdhuggingface/transformers/AGENTS.md
.ai/AGENTS.md
Hugging FaceAGENTS.mdhuggingface/transformers/.ai/AGENTS.md
Run make style (or make fix-repo) as the last step before opening a PR. Hosted review agents read this file via the root AGENTS.md / CLAUDE.md symlinks. Local agents wire their own assets: make codex (→ .agents/), make claude (→ .claude/). If your approach is materially different, say why a second PR is needed in the issue. - No one-off PRs for tiny edits (a single typo, an isolated lint fix). Mechanical cleanups are fine, but not as a first…
LangChainAGENTS.mdlangchain-ai/langchain/AGENTS.md
This document provides context to understand the LangChain Python project and assist with development. When Corridor's analyzePlan tool is available, create a plan and use the tool to analyze it before generating or modifying code. Apply the resulting security guidance before writing code. This is a Python monorepo with multiple independently versioned packages that use uv. This monorepo uses uv for dependency management. Local development uses editable installs declared under each package's [tool.uv.sources] table in pyproject.toml, so path dependencies resolve…
VercelAGENTS.mdvercel/next.js/AGENTS.md
This is a pnpm monorepo containing the Next.js framework and related packages. The main Next.js framework lives in packages/next/. This is what gets published as the next npm package. Source code is in packages/next/src/. Key entry points: Compiled output goes to packages/next/dist/ (mirrors src/ structure). Before editing or creating files in any subdirectory (e.g., packages/, crates/), read all README.md files in the directory path from the repo root up to and including the target file's directory. This helps identify any…
VercelAGENTS.mdvercel/next.js/evals/AGENTS.md
Do not export evals with "publish": false to next-evals-oss or publish their results on nextjs.org/evals.
VercelAGENTS.mdvercel/next.js/evals/next-upgrade/evals/future-cache-components/AGENTS.md
You may edit this application and create local commits. Do not push or create pull requests. Use npm. Preserve the behavior described in README.md.
VercelAGENTS.mdvercel/next.js/evals/next-upgrade/evals/future-cache-components-nudge/AGENTS.md
Do not push or create pull requests. Use npm.
VercelAGENTS.mdvercel/next.js/evals/next-upgrade/evals/latest-nudge/AGENTS.md
You may edit this application. Do not push or create pull requests. Use npm. Preserve the behavior described in the user prompt.
VercelAGENTS.mdvercel/next.js/.github/AGENTS.md
Set permissions: {} at the workflow level and grant the minimum needed per-job. Prefer GitHub-provided (actions/*) and Vercel-owned actions. For third-party actions: ${{ ... }} is interpolated into the script before bash runs, so untrusted values (PR title, branch name, issue body) can break out and execute. Route them through an env var and quote on use:
VercelAGENTS.mdvercel/next.js/packages/next/AGENTS.md
This version has breaking changes — APIs, conventions, and file structure may all differ from your training data. Read the relevant guide in dist/docs/ before writing any code. Heed deprecation notices.
VercelAGENTS.mdvercel/next.js/test/AGENTS.md
No summary in the file. Open it to read it.
An open format for instructions to coding agents, read by Codex, Cursor and others. Think of it as a README written for agents.
At the repository root, with more specific files in subdirectories. Agents read the one closest to the file they're editing.
Setup and test commands, code style, and the rules a new contributor would need to know.
Claude Code reads CLAUDE.md. A one-line CLAUDE.md that points at AGENTS.md covers both.
