react-ts-vite
AGENTS.md1 starsChanged 25 days agoMITView on GitHub
.agent/ makes the project legible to AI coding agents — Claude Code, OpenAI Codex, Cursor, GitHub Copilot, Gemini CLI, Jules, Amp and anything else that reads AGENTS.md. It is also the best onboarding document for humans. Everything is plain Markdown under version control; nothing is hidden in a vendor dashboard. Skills follow the Agent Skills specification: the folder name equals name, description says what the skill does and when to use it (≤ 1024 characters), optional license, compatibility, metadata and…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/rtv-architecture.mdcCursor rule
- .cursor/rules/rtv-code-style.mdcCursor rule
- .cursor/rules/rtv-design-system.mdcCursor rule
- .cursor/rules/rtv-git-workflow.mdcCursor rule
- .cursor/rules/rtv-testing.mdcCursor rule
- accessibility-auditSkill
- ci-cd-dockerSkill
- clean-code-reviewSkill
- dependency-upgradeSkill
- deployment-insightsSkill
- project-architectureSkill
- react-componentSkill
- scaffold-featureSkill
- tailwind-design-systemSkill
- testing-vitestSkill
The file
We split this agents.md into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 56
Rule: add .agent/rules/<topic>.md with front matter title and description.
- Very specific: line 59
Run npm run agent:sync and commit the regenerated files together with your change.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AI agents: the .agent/ folder
.agent/ makes the project legible to AI coding agents — Claude Code, OpenAI Codex, Cursor, GitHub Copilot, Gemini CLI, Jules, Amp and anything else that reads AGENTS.md. It is also the best onboarding document for humans. Everything is plain Markdown under version control; nothing is hidden in a vendor dashboard.
.agent/
├── README.md How the folder works
├── instructions.md Project-wide instructions: mindset, how to work, never-do list, scripts
├── rules/ Always-on rules
│ ├── architecture.md GENERATED from rtv.config.json — active architecture, layers, aliases, import rules
│ ├── code-style.md TypeScript/React/module conventions
│ ├── design-system.md Tailwind tokens, dark mode, Headless UI, Heroicons, accessibility
│ ├── testing.md Vitest + Testing Library conventions
│ └── git-workflow.md Conventional Commits, PRs, releases
├── skills/ Reusable know-how (Agent Skills standard: SKILL.md with YAML front matter)
├── workflows/ Step-by-step procedures (become slash commands in Claude Code)
└── agent.json GENERATED indexSkills
| Skill | Use it when… |
project-architecture | deciding where code lives, fixing arch:check violations, switching architectures |
react-component | building, extracting or refactoring components, hooks and pages |
scaffold-feature | implementing a business capability end to end ($ARGUMENTS = feature name) |
tailwind-design-system | styling, re-theming, adding dialogs/menus/listboxes |
testing-vitest | writing or fixing tests, raising coverage |
clean-code-review | reviewing a diff or PR against the project rules |
accessibility-audit | checking semantics, keyboard, focus, contrast, forms |
dependency-upgrade | bumping Vite/React/Tailwind/Vitest/TypeScript safely |
ci-cd-docker | debugging CI, adding workflows or deploy targets, running containers |
deployment-insights | producing release notes / risk reports, configuring Claude on deployment |
Skills follow the Agent Skills specification: the folder name equals name, description says what the skill does and when to use it (≤ 1024 characters), optional license, compatibility, metadata and allowed-tools. Only spec fields are used so the skills stay portable across tools.
How each tool consumes it
npm run agent:sync (also run by postinstall, by npm run setup, by the architecture switcher and by the Claude Code hook) renders the folder into tool-specific locations. npm run agent:check (part of npm run check) fails when any of them is stale.
| Tool | Reads | Notes |
| Claude Code | CLAUDE.md → @AGENTS.md import, .claude/skills/<skill>/ (mirrors), .claude/commands/<workflow>.md (mirrors), .claude/settings.json | Skills are invoked as /project-architecture, workflows as /add-page, etc. The settings pre-approve npm run … scripts and run Prettier after every edit (scripts/hooks/post-edit.mjs, which also re-syncs when .agent/ changes). |
| Codex, Jules, Amp, Zed, Warp, Gemini CLI, … | AGENTS.md | Generated: instructions + all rules inline + skill and workflow index with links. |
| GitHub Copilot | .github/copilot-instructions.md | Generated: instructions + rules + skill list. |
| Cursor | .cursor/rules/rtv-*.mdc | One always-apply rule per .agent/rules/*.md. |
Why mirrors instead of symlinks? They work on every OS and in every tool, and the check keeps them honest. Edit the source in .agent/; never edit a mirror.
The generated architecture rule
.agent/rules/architecture.md is rendered from rtv.config.json and the architecture catalogue every time you run npm run arch -- apply, npm run templates -- add/remove, npm run setup or npm run agent:sync. It tells agents the active architecture, where each role lives, the aliases, the layer table with allowed imports, and the applied templates — so an agent never has to guess where a file belongs.
Adding or changing instructions
- Rule: add
.agent/rules/<topic>.mdwith front mattertitleanddescription; keep it short and imperative. - Skill:
mkdir .agent/skills/<name>+SKILL.md(front mattername,description); add supporting files next to it (the deployment-insights skill ships itsPROMPT.md). - Workflow:
.agent/workflows/<name>.mdwith numbered steps referencing skills and scripts. - Run
npm run agent:syncand commit the regenerated files together with your change.
Design notes
- The folder is tool-agnostic on purpose: a team can switch agents without rewriting its conventions.
- Rules are the short, always-on layer; skills carry the long, situational detail; workflows chain them. Keep each rule under ~40 lines and each
SKILL.mdunder ~500. - Generated artefacts carry an
<!-- rtv:generated -->marker so the tooling never overwrites a hand-written file.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.