agentleFS
dimitar-radenkovdimitar-radenkov/Pointframe/website/llms.txt

Pointframe / website

llms.txt15 starsChanged 24 days agoMITView on GitHub

Pointframe lets your coding agent test the real Windows UI. It is a free, open-source (MIT) toolkit, Windows x64 only: agents inspect your app, interact with its controls, and run repeatable checks with screenshots and signed evidence (pointframe verify run). Its standalone MCP server and CLI also let AI agents list monitors and windows, capture screenshots, read on-screen text with Windows OCR, search saved captures, and record redacted MP4s; the same project ships a screenshot and recording tray app for…

The file

We split this llms.txt into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is 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.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run everything from the project root, in an interactive, unlocked Windows desktop session.

    line 10
  2. Distinctive:

    Write the spec, with your built exe's path relative to the project root (one line, for a single-project WPF app):

    line 19
  3. Very specific:

    Add one scenario to the scenarios array of .pointframe/verify.json: an action, the resulting state, and a negative control (a check with …

    line 25

≈ “Common” is an estimate: at least 1 in 5 of the 3,592 llms.txt 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
llms.txt
# Pointframe

> Pointframe lets your coding agent test the real Windows UI. It is a free, open-source (MIT) toolkit, Windows x64 only: agents inspect your app, interact with its controls, and run repeatable checks with screenshots and signed evidence (`pointframe verify run`). Its standalone MCP server and CLI also let AI agents list monitors and windows, capture screenshots, read on-screen text with Windows OCR, search saved captures, and record redacted MP4s; the same project ships a screenshot and recording tray app for people. Everything runs locally; official builds of the MCP server and the CLI send only anonymous usage counts (tool or command name, outcome, duration bucket, host, client type, version, random install ID), and POINTFRAME_TELEMETRY_OPTOUT=1 or DO_NOT_TRACK=1 turns that off. Windows 10 and 11, interactive desktop session required.

Install the MCP server for Claude Desktop (open the .mcpb bundle), Claude Code (`claude mcp add`), Codex (`codex mcp add`), VS Code, or Cursor. The exact commands are on the agent page below and in the MCP reference.

## Short path for coding agents

<!-- short-path:begin (generated from Pointframe.Mcp/Guides/verify-desktop-work.md by scripts/sync-agent-guides.ps1; do not edit) -->
Run everything from the project root, in an interactive, unlocked Windows desktop session.

1. Install the CLI (skip this if `pointframe --version` works). Run this one line as it is, in one tool call; it works from Bash and from PowerShell:

```
pwsh -NoProfile -Command '$d=Join-Path $env:TEMP "pointframe-cli"; Remove-Item $d -Recurse -Force -ErrorAction Ignore; New-Item -ItemType Directory $d | Out-Null; try { $b=Invoke-RestMethod "https://api.github.com/repos/dimitar-radenkov/Pointframe/releases?per_page=100" -Headers @{"User-Agent"="pointframe-install";"Accept"="application/vnd.github+json"} | Where-Object { $t=[string]$_.tag_name; $n=$t.Substring(1); $a=@($_.assets | ForEach-Object { $_.name }); -not $_.draft -and -not $_.prerelease -and $_.published_at -and $t -cmatch "^v\d+\.\d+\.\d+$" -and @("Pointframe.Cli-$n-win-x64.zip","Pointframe.Cli-$n-win-x64.zip.sha256","Pointframe.Mcp-$n-win-x64.mcpb" | Where-Object { $a -cnotcontains $_ }).Count -eq 0 } | Sort-Object { [version]$_.tag_name.Substring(1) } -Descending | Select-Object -First 1; if (-not $b) { throw "no complete agent release on the first page" }; $v=$b.tag_name.Substring(1); $u="https://github.com/dimitar-radenkov/Pointframe/releases/download/v$v/Pointframe.Cli-$v-win-x64.zip" } catch { Write-Warning "Could not find the newest agent release ($($_.Exception.Message)); using the legacy latest-release download."; $u="https://github.com/dimitar-radenkov/Pointframe/releases/latest/download/Pointframe.Cli-win-x64.zip" }; Invoke-WebRequest $u -OutFile "$d\cli.zip"; Invoke-WebRequest "$u.sha256" -OutFile "$d\cli.zip.sha256"; if ((Get-FileHash "$d\cli.zip").Hash -ne ((Get-Content "$d\cli.zip.sha256" -Raw).Trim() -split "\s+")[0]) { throw "SHA-256 mismatch" }; Expand-Archive "$d\cli.zip" "$d\x"; & "$d\x\Pointframe.Cli.exe" install; "$env:LOCALAPPDATA\Programs\Pointframe.Cli\pointframe.exe"'
```

   The line takes the newest release that has the CLI zip and its checksum from the GitHub Releases API, and prints a warning if it has to use the latest-release download instead. `install` puts `pointframe` on the user `Path` for new shells only. In the shell you already have open, call the exe path the line prints (`%LOCALAPPDATA%\Programs\Pointframe.Cli\pointframe.exe`) in place of `pointframe`.
2. Write the spec, with your built exe's path relative to the project root (one line, for a single-project WPF app):

```
pointframe verify init --app bin/Release/net10.0-windows/MyApp.exe --hooks none
```

3. Add one scenario to the `scenarios` array of `.pointframe/verify.json`: an action, the resulting state, and a negative control (a check with `expectFailure` and a deliberately wrong expectation, which must fail):

```jsonc
{ "id": "save-text", "criteria": ["Saving shows the typed text in the status line."], "steps": [
  { "enterText": { "automationId": "textBox", "text": "hello" } },
  { "invoke": { "automationId": "saveButton" } },
  { "check": { "kind": "textEquals", "automationId": "statusLabel", "expected": "Saved hello", "criterion": "C1" } },
  { "check": { "kind": "textEquals", "automationId": "statusLabel", "expected": "Not saved", "expectFailure": true, "timeoutSeconds": 1 } } ] }
```

4. Make it a `verify-<app>` skill (pstack-compatible), so the next agent can repeat the proof. Use ONE folder, the project skill folder the repo already has: `.claude/skills/` for Claude Code, `.agents/skills/` for Codex and other agents. The folder and the `name:` in the SKILL.md frontmatter are both `verify-<app>`:

```
<skills>/verify-<app>/SKILL.md              (frontmatter: name: verify-<app>, description: Prove that a change to <app> works ... Use after changing <app>.)
<skills>/verify-<app>/features/README.md    (lists the feature files)
<skills>/verify-<app>/features/<feature>.md (one per feature)
```

   Each feature file has exactly four H2 sections, with these exact headings: `Sub-features`, `How to get to it (user POV)`, `Driving it with Pointframe`, `Gotchas`. `Driving it with Pointframe` and `Gotchas` name the scenario id (for example `save-text`) and say to run `pointframe verify run --scenario <id>`; they never copy steps from the spec. Write all skill files in one shell command, before step 5.
5. Run `pointframe verify run`. It runs the gates, launches the app, drives the scenario in the real UI, and installs the MCP server itself on its first run. Read `status` in the printed JSON: `pass` is exit 0, `fail` is exit 1, a wrong spec is exit 2. Run `verify run` last: any later edit, even to docs or skill files, makes the pass stale (`verify status` then shows `fresh: false`), so write everything first and run again if you change anything.
6. Evidence: `artifacts/pointframe-verify/verdict.json` is the verdict, and each scenario's signed bundle (`report.json`, `evidence/` screenshots, `index.html`) is under `artifacts/pointframe-verify/runs/<time>/`. Say the work is done only when `status` is `pass`.

The spec fields are in [verify-spec.md](https://github.com/dimitar-radenkov/Pointframe/blob/master/docs/cli/verify-spec.md); the example above covers the common case.
<!-- short-path:end -->

## Agents

- [Windows screenshot MCP server for AI agents](https://dimitar-radenkov.github.io/Pointframe/windows-screenshot-mcp-server.html): what the server does, install steps per client, the eleven tools, and the privacy model
- [Let your coding agent test the real Windows UI](https://dimitar-radenkov.github.io/Pointframe/verify-coding-agents-windows.html): inspect your app, drive its controls, and run repeatable checks with screenshots and signed evidence; build and test gates and a Stop hook included
- [Coding agents: verify a Windows desktop app in six steps](https://github.com/dimitar-radenkov/Pointframe/blob/master/docs/cli/verify.md#coding-agents-start-here)

## Documentation

- [MCP reference](https://github.com/dimitar-radenkov/Pointframe/blob/master/docs/mcp/README.md): tools, resources, workflow, and per-client install
- [Privacy policy](https://dimitar-radenkov.github.io/Pointframe/privacy.html): what the desktop app, CLI, and MCP server collect, where it goes, retention, and the opt-out
- [PRIVACY.md](https://github.com/dimitar-radenkov/Pointframe/blob/master/PRIVACY.md#privacy-policy): the same policy in the repository
- [CLI guide](https://github.com/dimitar-radenkov/Pointframe/blob/master/docs/cli/README.md): commands, JSON output, exit codes, and the MCP installer
- [Desktop-testing MCP guide](https://github.com/dimitar-radenkov/Pointframe/blob/master/docs/mcp-desktop-testing/README.md): the opt-in tools that drive a Windows app, off by default
- [pointframe verify](https://github.com/dimitar-radenkov/Pointframe/blob/master/docs/cli/verify.md): declarative done-checks for coding agents

## Releases

- [Latest release](https://github.com/dimitar-radenkov/Pointframe/releases/latest): installer, CLI ZIP, MCP bundle, and checksums
- [Pointframe.Mcp-win-x64.mcpb](https://github.com/dimitar-radenkov/Pointframe/releases/latest/download/Pointframe.Mcp-win-x64.mcpb): the MCP server bundle (self-contained, includes ffmpeg)
- [Pointframe.Mcp-win-x64.mcpb.sha256](https://github.com/dimitar-radenkov/Pointframe/releases/latest/download/Pointframe.Mcp-win-x64.mcpb.sha256): its SHA-256
- [Pointframe.Cli-win-x64.zip](https://github.com/dimitar-radenkov/Pointframe/releases/latest/download/Pointframe.Cli-win-x64.zip): the standalone CLI
- [MCP Registry entry](https://registry.modelcontextprotocol.io/v0/servers?search=pointframe): io.github.dimitar-radenkov/pointframe-mcp

## Optional

- [Source code](https://github.com/dimitar-radenkov/Pointframe)
- [Pointframe website](https://dimitar-radenkov.github.io/Pointframe/)

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.