How real projects brief Codex, Cursor and every other agent that reads AGENTS.md.
OpenAIAGENTS.mdopenai/planttalk/AGENTS.md
This repo is designed to be opened directly in Codex Desktop. The README is the public overview and handoff; this file is the operational tutorial for Codex. Do not ask users to install a separate Codex skill. Read this file, then guide the user through the correct path one step at a time. - Act like the tutorial guide, not like a docs search engine. - Guide one step at a time. Do not dump the whole build plan unless…
OpenAIAGENTS.mdopenai/openai-guardrails-python/AGENTS.md
This guide defines the required workflow for agents and contributors working in the OpenAI Guardrails Python repository. Repository skills live under .agents/skills/. A reference such as $<skill-name> is a repository instruction reference, not a request for manual user invocation. When a rule requires a skill, read .agents/skills/<skill-name>/SKILL.md completely before taking task actions, follow it, and resolve referenced files relative to that skill directory. This repository defines these repository skills: Use $final-release-review when the user asks for pre-release planning or a…
OpenAIAGENTS.mdopenai/fence/AGENTS.md
This repository is the Rust implementation scaffold for Fence, a security agent intended to harden a narrowly supported CI runner class. The v0 protected target is GitHub-hosted Ubuntu LTS x64, selected through either ubuntu-24.04 or ubuntu-latest, with a native Linux GNU agent artifact; see docs/v0.md for the normative behavior contract and docs/threat-model.md for security claims and residual risks. It is bootstrapped from a hermetic, reproducible, air-gapped-leaning Rust template. Treat it as public open source infrastructure: every file, comment, workflow, log…
OpenAIAGENTS.mdopenai/snap-o/AGENTS.md
Snap-O is a public, open-source Android inspection tool. Use synthetic data in tests and examples. Never publish credentials, captured private traffic, confidential implementation details, or private issue-tracker links, identifiers, or content. Review the full diff, commit message, and PR text for sensitive information before publishing. Follow release/README.md when changing protocols, published APIs, version metadata, or packaging. Update release/ checks and tests when protocol definitions or source paths change. Use plain technical English in all prose, including READMEs, docs, code comments,…
OpenAIAGENTS.mdopenai/snap-o/app-macos/AGENTS.md
For automated contributors:
OpenAIAGENTS.mdopenai/snap-o/tools/AGENTS.md
This is the code repository for the Snap-O macOS app and optional Android libraries to help provide data to Snap-O.
OpenAIAGENTS.mdopenai/snap-o/tool-sdk/host/AGENTS.md
No summary in the file. Open it to read it.
OpenAIAGENTS.mdopenai/openai-guardrails-js/AGENTS.md
When submitting a PR, always monitor CI for failures. Automatically fix and re-push only failures introduced or worsened by the change, or narrowly necessary to achieve the requested outcome. Report unrelated preexisting failures and retry transient or flaky checks when appropriate; do not expand the diff to fix them. Once CI passes, post in #sdk-reviews and ask for a review. Always post in the root #sdk-reviews channel, do not post in threads. When addressing feedback on a PR - always…
OpenAIAGENTS.mdopenai/birdingpal/AGENTS.md
This repository is a project recap for Birding Pal, not a step-by-step public build tutorial. When editing this repo: - Preserve the README's "what we did" tone. - Keep Android details in supporting files unless they are necessary to explain the technical implementation. - Do not change the firmware wiring diagram or schematic unless explicitly asked. - Do not add secrets, API keys, Wi-Fi credentials, private certificates, personal data, unapproved media, datasets, model weights, or generated build artifacts. - Keep…
OpenAIAGENTS.mdopenai/redcard/AGENTS.md
This repo is designed to be opened in the ChatGPT macOS app and operated through Codex. The normal Codex activation path is the repo-owned $redcard skill. Red Card watches for a physical red card through the webcam. When detected during a Google Meet call, it runs the full sequence: The default entrypoint is: The default config is: The README is the public handoff. Once the user asks for help setting up, running, or debugging Red Card, Codex should become the…
OpenAIAGENTS.mdopenai/terraform-provider-openai/AGENTS.md
Read CONTRIBUTING.md, SECURITY.md, and RELEASING.md before changing this repository. - The provider manages OpenAI Administration API resources with privileged organization credentials and beta project webhook resources with project credentials. Treat authentication, authorization, Terraform state, and published provider artifacts as security-sensitive boundaries. - Check .terraform-generator-manifest.json before editing provider sources, generated tests, Terraform examples, or docs/. Files listed in the manifest and files marked Code generated ... DO NOT EDIT are generator-owned; change their authoritative upstream source when available instead of introducing…
OpenAIAGENTS.mdopenai/homebrew-tools/AGENTS.md
This is the openai/tools Homebrew tap. It distributes generated Ruby recipes for prebuilt tools; it does not contain their application source or generators. Read CONTRIBUTING.md and SECURITY.md before making changes. The recipe and upstream inventory is in Generated recipes. - Before editing, state the requested outcome, acceptance criteria, affected paths, and non-goals. Keep changes limited to that outcome; ask before changing public interfaces, crossing ownership boundaries, or restructuring architecture. - Use a linked Git worktree, verify its intended base commit,…
OpenAIAGENTS.mdopenai/community-plugins/AGENTS.md
These conventions apply throughout this repository and reflect the maintainer's preferred organization. Follow the existing Grafana, ReviewOps and React Native to SwiftUI plugins when adding or changing a plugin. - Keep each installable package in plugins/<plugin-name>/. Repository companion tests belong in tests/<plugin-name>/, with unit/, integration/, security/ and support/ directories where applicable. Existing deeper package suites should remain connected to the root master command. - Provide one canonical root npm run test:<plugin-name> command that runs the complete automated suite: package, functional,…
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.
