Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add trsdn/mcp-server-uiautomation --skill uia-mcpgit clone --depth 1 https://github.com/trsdn/mcp-server-uiautomationWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/trsdn/mcp-server-uiautomation/uia-mcp)<a href="https://agentmods.dev/skills/trsdn/mcp-server-uiautomation/uia-mcp"><img src="https://agentmods.dev/badge/skills/trsdn/mcp-server-uiautomation/uia-mcp/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/trsdn/mcp-server-uiautomation/uia-mcp"><img src="https://agentmods.dev/badge/skills/trsdn/mcp-server-uiautomation/uia-mcp.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00064 | $0.00926 |
| Opus 5 | $0.00032 | $0.00463 |
| Sonnet 5 | $0.00013 | $0.00185 |
| Haiku 4.5 | $0.00006 | $0.00093 |
Grade A, and why
uia-mcp scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Use the UIAutomationMcp MCP server to see and drive real Windows desktop UI. Every tool returns JSON.
Tools
| Tool | Purpose |
|---|---|
uia_desktop |
Probe the desktop root; confirms UI Automation is reachable |
uia_snapshot |
Desktop root plus focused element in one call |
uia_focused |
The element with keyboard focus |
uia_handle |
Element owning a native window handle |
uia_point |
Element at a screen point |
uia_find_name / uia_find_class / uia_find_automation_id |
Single-criterion shortcuts |
uia_inspect |
Full metadata for one element, using the generic locator |
uia_find |
Multiple matches, with scope and result cap |
uia_children / uia_descendants |
Enumerate from raw, control, or content view |
uia_navigate |
Parent, first/last child, sibling, normalize |
uia_text |
Document text, selection, caret, annotations, IME state, TextChild |
uia_selection |
Selection container state |
uia_table |
Read a Grid/Table control as a cell matrix with headers |
uia_action |
Invoke, focus, toggle, expand, scroll, move, resize, and more |
uia_wait_event |
One-shot wait for a UI Automation event |
The locator model
Most tools share the same parameters: root, focused, fromFocused,
handle, x/y, name, className, automationId, frameworkId,
controlType, processId, scope, plus the negative forms notName,
notClassName, notAutomationId, notControlType. They AND-compose.
Prefer uia_table over walking cells with uia_descendants, and prefer a
narrow locator over fetching a large result set and filtering.
uia_inspect accepts tryInspect: true to return null instead of failing
when nothing matches — use it to assert that something is gone, such as a
dialog that should have closed.
Two behaviours that surprise people
Virtualized items. A virtualizing container only materializes what it is
showing, so a 300-file folder may expose a dozen live elements. uia_inspect,
uia_find and uia_action automatically fall back to asking ItemContainer
providers and realizing the result, so a plain locator usually just works. Pass
realizeVirtualized: false when you specifically want to prove absence.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago Changed · +66 lines · +61 tokens per session 6e040d89d749
- 11d ago First seen · 12 lines · 3 tokens per session scan A 14b62e34d788
uia-mcp is a skill published in the GitHub repository trsdn/mcp-server-uiautomation (2 stars, last pushed 6d ago), licensed MIT. It adds 64 tokens to every session and 926 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
agent-desktop
Reliable computer use via native OS accessibility trees. Use when an AI agent needs to see and operate desktop applications (click buttons, fill forms, navigate menus, read UI state, toggle checkboxes, scroll, drag, type text, take screenshots, manage windows, use clipboard, manage notifications). Covers 59 command…
agent-desktop-ffi
C-ABI bindings over agent-desktop's PlatformAdapter. Consumers (Python ctypes, Swift, Node ffi-napi, Go cgo, C++, Ruby fiddle) link libagentdesktopffi.{dylib,so,dll} and call ad functions directly instead of spawning the CLI binary per call. The canonical observe-act workflow is: adinit → adadaptercreate[withsession]…
macos-harness
Control a whole Mac from one persistent Python session with screenshots, PID-targeted input, an animated virtual pointer, targeted Apple Accessibility, Apple Events, Browser Harness CDP, and filesystem access. Use for native, Electron, browser, dialog, file, or cross-app tasks without moving the physical cursor or…
axe-core Accessibility Automation
Automated accessibility testing with axe-core integrated into CI pipelines, including custom rule configuration, issue prioritization, and remediation guidance.
kmsg
Use KMSG to read, watch, and send KakaoTalk messages safely from a coding agent. Use whenever the user mentions KakaoTalk or kmsg and wants chat discovery, recent messages, monitoring, text or image delivery, setup checks, or CLI troubleshooting. Do not use for other messaging platforms or private KakaoTalk protocol…
agent-orchestration
A guide for coordinating independent AI coding agents such as Claude Code, OpenCode, or Hermes. It explains how to register each agent, check which ones are running, choose one for a task, and record stable information about how to use it.