How real projects brief Codex, Cursor and every other agent that reads AGENTS.md.
MicrosoftAGENTS.mdmicrosoft/mcp-for-beginners/translations/en/AGENTS.md
links for review ```bash # List Markdown links (pattern audit) find . -name "*.md" -not -path "*/node_modules/*" -not -path "./translations/*" -not -path "./translated_images/*" -print0 | xargs -0 grep -En "\[.*\]\(.*\)" ``` 2. **Code Sample Validation
MicrosoftAGENTS.mdmicrosoft/mcp-for-beginners/translations/pcm/AGENTS.md
links for review ```bash # List Markdown links (patin audit) find . -name "*.md" -not -path "*/node_modules/*" -not -path "./translations/*" -not -path "./translated_images/*" -print0 | xargs -0 grep -En "\[.*\]\(.*\)" ``` 2. **Code Sample Validation
LangChainAGENTS.mdlangchain-ai/openwiki/AGENTS.md
follow its links to the relevant pages. - Treat source code and tests as authoritative. A brief's unknowns and review items are verification gaps, not automatic requirements. - Prefer the narrowest
AppleAGENTS.mdapple/foundationdb/AGENTS.md
year from now? Is it genuinely essential and **non-obvious from the code itself**? ## Code Review Unless you have specific instructions to the contrary, when asked to review code (named
lidge-junAGENTS.mdlidge-jun/opencodex/AGENTS.md
This file summarizes; it never overrides it. ## Review guidelines These rules apply to all code reviews on this repository, including automated reviewers (Codex, CodeRabbit). - **Language:** always review in English, regardless
lidge-junAGENTS.mdlidge-jun/opencodex/.github/AGENTS.md
boundary Every workflow, ownership, branch-enforcement, release, or repository-automation change requires explicit security review under `MAINTAINERS.md`. ## Workflow rules - Grant the minimum required `permissions`. - Pin third-party actions to immutable
lidge-junAGENTS.mdlidge-jun/opencodex/gui/AGENTS.md
cache `c`/`w`) - URLs / localhost endpoints, adapter ids (`oauth`, `passthrough`, npm channels) - Keep **code comments** (including shell `# …` comments in samples). Never strip them to “satisfy” i18n. - Run `bun run lint
lidge-junAGENTS.mdlidge-jun/opencodex/src/AGENTS.md
APIs. Introduce a Node-only runtime dependency only when the task explicitly requires compatibility code and the owning module already has that role. - Preserve existing public exports and configuration compatibility
lidge-junAGENTS.mdlidge-jun/opencodex/structure/AGENTS.md
that followed it. Reasoning written as ordinary prose is not detectable and stays a review judgement. The check is a literal match, which is why this line describes the marker
earthtojakeAGENTS.mdearthtojake/text-to-cad/AGENTS.md
that covers the change. Use broad wrappers when touching shared surfaces or before handoff: - Code tests: `scripts/test/test.sh` (JS, then Python, then policy). - Focused runners: `scripts/test/test-js.sh`, `scripts/test/test-docs.sh`, `scripts/test/test-python.sh`, `scripts/test/test-global.sh`. `test-python.sh` takes
triggerdotdevAGENTS.mdtriggerdotdev/trigger.dev/AGENTS.md
namespace catalog, running crumbs init" load: "node_modules/agentcrumbs/skills/agentcrumbs/init/SKILL.md" ## agentcrumbs Add crumbs as you write code — not just when debugging. Mark lines with `// @crumbs` or wrap blocks in `// #region @crumbs`. They
Hugging FaceAGENTS.mdhuggingface/transformers.js/.ai/AGENTS.md
skills/transformers-js/SKILL.md) — how to use the library itself. Load this skill when working on code that calls `@huggingface/transformers`. ## Contributing Before opening a pull request: - **Check for existing work.** Search open
VercelAGENTS.mdvercel/vercel/packages/cli/AGENTS.md
design system lives in `packages/cli/.agents/skills/cli-ux/` and should be loaded only when the task or code review touches user-facing CLI behavior. ## First Steps 1. Inspect the current source and tests
HashiCorpAGENTS.mdhashicorp/packer/AGENTS.md
generate-check` so generated code stays in sync. - Do not leave formatting or generated-file drift for the user to discover at push or review time; run the relevant `make
AgriciDanielAGENTS.mdAgriciDaniel/claude-obsidian/AGENTS.md
package for building source-cited, compounding Obsidian knowledge bases. It also ships a Claude Code plugin adapter. The portable workflow is implemented in `skills/` and the standard-library `claude_obsidian
NanmiCoderAGENTS.mdNanmiCoder/cc-haha/AGENTS.md
need to run every lane separately first. Reuse passing results for unchanged code; rerun or broaden checks when subsequent edits, failures, or unresolved risks warrant it. - Required PR checks must
NVIDIAAGENTS.mdNVIDIA/TensorRT-LLM/AGENTS.md
Protected APIs exist** — changes to LLM API signatures will fail `tests/unittest/api_stability` tests. Get code owner review. - **Integration tests need GPUs + models** — always set `LLM_MODELS_ROOT` and ensure GPU access
NVIDIAAGENTS.mdNVIDIA/TensorRT-LLM/cpp/tensorrt_llm/batch_manager/kv_cache_manager_v2/AGENTS.md
error -- and the corruption appears in whatever code runs next, not at the mismatch. ## High-risk changes Use extra review and tests for changes involving: - destructor, `shutdown()`, `close()`, tree detachment
superset-shAGENTS.mdsuperset-sh/superset/AGENTS.md
Contributors are responsible for every change they submit, including agent-generated code. They must understand the changes, review the diff, and verify the behavior before requesting review. - Ask for clarification
palmier-ioAGENTS.mdpalmier-io/palmier-pro/AGENTS.md
verification plan. Do not substitute a narrower test while claiming end-to-end coverage. ## Code reviews - Review for correctness before style. Trace the changed behavior through its callers, shared state
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.
