daveebbelaardaveebbelaar/bawa/AGENTS.md

bawa

AGENTS.md17 starsChanged 10 days agoMITView on GitHub

Agent skills for building native Windows apps. Installed as a Claude Code and Codex plugin from .claude-plugin/, or through npx skills add. AGENTS.md at the root is a copy of this file (.claude/CLAUDE.md); keep them identical. BAWA is the Windows sibling of BAMA (Build Any Mac App). Same pipeline, same skill names where the job is the same, same writing rules. Only the stack differs. - C# 14 on .NET 10, WinUI 3 (Windows App SDK 2.x), XAML with x:Bind.…

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 31,843 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 repository23 other files it gives its agents.
- .claude/CLAUDE.mdCLAUDE.md
- agent-cliSkill
- csharp-winuiSkill
- improve-architectureSkill
- os-integrationsSkill
- performanceSkill
- app-iconSkill
- design-from-appSkill
- design-kitSkill
- design-systemSkill
- layout-gridSkill
- ux-designSkill
- windows-designSkill
- cloneSkill
- researchSkill
- user-storiesSkill
- build-and-verifySkill
- distributeSkill
- improve-loopSkill
- test-windows-appSkill
- app-briefSkill
- bawaSkill
- new-windows-appSkill

The file

We split this AGENTS.md into the rules it gives an agent

6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Very specific:
Follow Matt Pocock's writing-for-agents rules:
line 59
Distinctive:
Every step ends on a checkable condition.
line 66
Distinctive:
No em dashes. Plain words, short sentences, active voice.
line 73
Very specific:
Create skills/<stage>/<name>/SKILL.md and agents/openai.yaml.
line 80
Very specific:
Add it to plugin.json skills, the README table, and the router (skills/start/bawa/SKILL.md).
line 81
Very specific:
Keep version in plugin.json at 1.0.0 until the first release.
line 84

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

AGENTS.md

BAWA

Agent skills for building native Windows apps. Installed as a Claude Code and Codex plugin from .claude-plugin/, or through npx skills add. AGENTS.md at the root is a copy of this file (.claude/CLAUDE.md); keep them identical.

BAWA is the Windows sibling of BAMA (Build Any Mac App). Same pipeline, same skill names where the job is the same, same writing rules. Only the stack differs.

Stack

C# 14 on .NET 10, WinUI 3 (Windows App SDK 2.x), XAML with x:Bind.
CommunityToolkit.Mvvm for observable models, CommunityToolkit.WinUI SettingsControls for the Settings page.
SQLite through Microsoft.Data.Sqlite in a <App>.Core library (no WinUI), tested with xUnit v3.
Unpackaged: a plain .exe with the Windows App SDK runtime in the app folder. No MSIX, no Store, no Visual Studio. The .NET SDK is the only install (winget install Microsoft.DotNet.SDK.10).
Scripts are PowerShell that runs in Windows PowerShell 5.1 and in pwsh 7.
Windows 11 first (Mica, the TitleBar control, rounded corners), Windows 10 19041 or later still runs.

Commands in a BAWA app

Job	Command
Build and install to `%LOCALAPPDATA%\Programs\<App>` plus a Start menu entry	`./build.ps1` (`-Dev` debug, `-Open` start it)
Unit tests	`dotnet test tests/<App>.Core.Tests`
Self-test on a seeded `%TEMP%` store	`tools/selftest.ps1`
Snapshots of every page in light and dark	`tools/snap.ps1` (`-Pages all,settings`, `-Accent green`, `-Grid`)
Grid checks	`tools/grid-lint.ps1`, `tools/check-layout.ps1`
Icon from `Icon/AppIcon.svg`	`tools/make-icon.ps1 -Review`
A build for other PCs (no .NET needed), checked by the self-test	`tools/publish.ps1` (`-Aot` NativeAOT, `-Zip`)

Environment flags the app reads: APP_STORE (a file in %TEMP%), APP_SEED=1, APP_SNAPSHOT=<dir>, APP_SNAPSHOT_PAGES, APP_ACCENT, APP_GRID=1, APP_LOG=<file>, and the --selftest argument. The tools run the installed app, or the build in APP_EXE.

Layout

skills/<stage>/<name>/SKILL.md, stages in pipeline order: start, plan, code, design, ship.
References sit next to SKILL.md (recipes.md, patterns.md) and are linked relatively. Scripts go in scripts/, templates in template/.
agents/openai.yaml in every skill: Codex display name and short description. A user-invoked skill also sets policy.allow_implicit_invocation: false.
.claude-plugin/plugin.json lists every shipped skill in skills.
specs/<app>.md: build plans for small apps, the Windows versions of BAMA's specs. Each has Brief, How it works, Stories, Gotchas, and a "Done when" line, in that order. specs/README.md lists them.

Writing skills

Follow Matt Pocock's writing-for-agents rules:

The description is a context pointer: one short sentence on what it is, then "Use when" and one trigger per branch. Front-load the leading word.
Inline what every branch needs. Push what only some branches need into a sibling file behind a link.
Every step ends on a checkable condition. Every skill ends with a "Done when" line.
Prompt the positive. Use a prohibition only as a hard guardrail, next to the positive target.
Cache only what the environment cannot tell the agent: conventions, reasons, gotchas from real apps.
One skill calls another with "Call the Skill tool with <name>".
No em dashes. Plain words, short sentences, active voice.
Windows words: "epx" (effective pixels) for sizes, Ctrl not Command, "the taskbar" and "Start" not "the Dock", sentence case for UI labels ("New item", not "New Item").

Adding a skill

Create skills/<stage>/<name>/SKILL.md and agents/openai.yaml.
Add it to plugin.json skills, the README table, and the router (skills/start/bawa/SKILL.md).
Run claude plugin validate . --strict.
Keep version in plugin.json at 1.0.0 until the first release. After that, bump it, so installed copies update.

Download now (ZIP)

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

What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free