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 TheSyart/emperor-agent --skill computer-usegit clone --depth 1 https://github.com/TheSyart/emperor-agentWrote 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/thesyart/emperor-agent/computer-use)<a href="https://agentmods.dev/skills/thesyart/emperor-agent/computer-use"><img src="https://agentmods.dev/badge/skills/thesyart/emperor-agent/computer-use/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/thesyart/emperor-agent/computer-use"><img src="https://agentmods.dev/badge/skills/thesyart/emperor-agent/computer-use.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 | $0.00099 | $0.03125 |
| Opus 5.5 | $0.00040 | $0.01250 |
| Sonnet 5.5 | $0.00020 | $0.00625 |
| Haiku 4.5 | $0.00010 | $0.00313 |
Grade A, and why
computer-use 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 3d 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 — 98 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Computer Use
How to work in web pages through Emperor's built-in browser and in macOS app windows. You drive the tabs and windows yourself from start to finish; the user watches them in a read-only preview in the conversation (desktop windows also show the system's purple screen-sharing mark) and does not pause or take them over.
Before you start
- Is the GUI needed? Prefer an API, a CLI,
web_fetchor a file tool when one can do the job. Use the browser only for the part that must happen in a page. - Are the tools there? This Skill is listed only while computer use is on. If the
browser_*tools disappear during a task, the user switched it off: stop and say so; do not try to drive a browser another way. They can turn it back on in 设置 › 电脑操作 or with the Computer Use plugin. - What can the driver do right now? Call
ui_get_capabilitiesonce when unsure. It lists the available drivers, their actions and current limitations. Cross-origin iframe observation and clicking work; typing into one is disabled.
The loop: observe → act → verify
browser_openthe page (one tab per task is usually enough).browser_observeto get a semantic snapshot. Elements carry refs such asr3.12.- Act on refs from the latest observation:
browser_fill,browser_click,browser_select,browser_press,browser_scroll. - Observe again and check the result the user asked for (the confirmation text, the new URL, the filled value) — not just that an action was sent.
- Use
browser_waitfor things that load later (text appears, URL changes, an element becomes enabled) instead of observing in a tight loop.
Rules that keep this reliable:
- Refs are valid only for the observation that produced them. After a navigation or a new observation, use the new refs; a
STALE_ELEMENTerror means observe again,STALE_TARGETmeans the tab or page you meant is gone — except with reasonreleased-at-task-end, which means an earlier task released it and it is still open (see "Between tasks"). - Several actions may run on one snapshot (fill, fill, click), as long as the page did not navigate in between.
- A click on something covered by an overlay is refused before anything is sent ("covered by another element"): close the overlay (cookie banner, dialog) first.
- Long pages are cut to a budget. Narrow
browser_observewithroleornameContains, usediff: trueto see only what changed, and pass the returnedcursorto read the rest — rather than scrolling blindly. browser_screenshotis for the user or for a vision-capable model; the semantic snapshot is the primary source of truth.
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.
- 3d ago Changed · +22 lines · +10 tokens per session e6d0a8a5e6d8
- 19d ago First seen · 76 lines · 89 tokens per session scan A caa74ec2c34e
computer-use is a skill published in the GitHub repository TheSyart/emperor-agent (199 stars, last pushed 4d ago), licensed MIT. It adds 99 tokens to every session and 3,125 once invoked, about $0.0004 per session on Opus 5.5. 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 2026-09-30.
Other skills, from other repositories
knowledge-wiki
Manage the personal knowledge wiki. Use when the user shares articles, documents, or asks to organize knowledge; when a conversation produces insights worth preserving as structured knowledge; or when the user asks about the knowledge base.
browserwing-executor
Control browser automation through HTTP API. Supports page navigation, element interaction (click, type, select), data extraction, accessibility snapshot analysis, screenshot, JavaScript execution, and batch operations.
dev-browser
Browser automation with persistent page state. Use when users ask to navigate websites, fill forms, take screenshots, extract web data, test web apps, or automate browser workflows. Trigger phrases include "go to [url]", "click on", "fill out the form", "take a screenshot", "scrape", "automate", "test the website"…
browserwing-admin
Manage and operate BrowserWing — an intelligent browser automation platform. Install dependencies, configure LLM, create/manage/execute automation scripts, use AI-driven exploration to generate scripts, browse the script marketplace, and troubleshoot issues.
bilibili-play
A browser-control skill for opening and playing videos on Bilibili, a Chinese video-sharing website, including selected episodes, playback speed, fullscreen mode, and frame capture.
web-access
Use this skill for live web research, webpage fetching, login-state browsing, dynamic pages, social sites, and tasks that need current internet evidence.