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 agentmods add skills/latestaiagents/agent-skills/computer-usenpx skills add latestaiagents/agent-skills --skill computer-usegit clone --depth 1 https://github.com/latestaiagents/agent-skillsWrote 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/latestaiagents/agent-skills/computer-use)<a href="https://agentmods.dev/skills/latestaiagents/agent-skills/computer-use"><img src="https://agentmods.dev/badge/skills/latestaiagents/agent-skills/computer-use.svg" alt="Measured on agentmods" 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 | $0.00100 | $0.01480 |
| Opus 5 | $0.00050 | $0.00740 |
| Sonnet 5 | $0.00020 | $0.00296 |
| Haiku 4.5 | $0.00010 | $0.00148 |
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 2d 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 — 153 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Computer Use
Computer Use lets Claude control a virtual computer — take screenshots, move cursor, click, type. Use it when there's no API, not as a first resort.
When to Use
- Automating legacy apps with no API
- Cross-app workflows (copy from A, paste to B, click approve)
- QA/UI testing where you want the agent to drive a real browser
- End-to-end automation mixing web and desktop
When NOT to Use
- Anything that has an API — APIs are 10-100× faster and cheaper
- High-frequency tasks — each action is a model call (slow)
- Anything where a failed click has real consequences (payments, deletes) without human in the loop
- Untrusted environments — the agent can be injected via screen content
Tool Shape
const tools = [
{ type: "computer_20250124", name: "computer", display_width_px: 1280, display_height_px: 800, display_number: 1 },
{ type: "text_editor_20250124", name: "str_replace_editor" }, // optional
{ type: "bash_20250124", name: "bash" }, // optional
];
const response = await client.beta.messages.create(
{
model: "claude-sonnet-4-6",
max_tokens: 4096,
tools,
messages: [{ role: "user", content: "Open the browser and find Claude's API pricing page." }],
},
{ headers: { "anthropic-beta": "computer-use-2025-01-24" } },
);
The model returns tool_use blocks with actions like screenshot, left_click, type, key, mouse_move, scroll.
Reference Implementation
Anthropic ships a reference Docker container (ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest) with:
- Virtual X server (Xvfb)
- Firefox + common apps
- VNC for human observation
- HTTP shim that executes tool calls
Don't run this on a host with real user data. It's a demo. For production, use a sandboxed cloud VM (Firecracker, Vercel Sandbox, cloud-hypervisor).
Action Loop
async function run(goal: string) {
const messages = [{ role: "user", content: goal }];
while (true) {
const response = await client.beta.messages.create({
model: "claude-sonnet-4-6",
max_tokens: 4096,
tools,
messages,
}, { headers: { "anthropic-beta": "computer-use-2025-01-24" } });
messages.push({ role: "assistant", content: response.content });
if (response.stop_reason !== "tool_use") return response;
const results = [];
for (const block of response.content) {
if (block.type === "tool_use") {
const output = await executeAction(block.name, block.input); // your VM controller
results.push({
type: "tool_result",
tool_use_id: block.id,
content: output.screenshot ? [{ type: "image", source: output.screenshot }] : output.text,
});
}
}
messages.push({ role: "user", content: results });
}
}
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.
- 2d ago First seen · 153 lines · 100 tokens per session scan A 40e8001286c6
computer-use is a skill published in the GitHub repository latestaiagents/agent-skills (5 stars, last pushed 4mo ago), licensed MIT. It adds 100 tokens to every session and 1,480 once invoked, about $0.0005 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 2026-09-03.
Other skills, from other repositories
ai-browser-use
业务诊断助手适合市场营销、运营、产品、销售在用户提出“这件事该怎么做”这类问题,需要快速拆解目标、判断重点并形成可执行结果时使用,帮助基于输入材料生成摘要、诊断结论、行动建议和可复用交付物。.
yutori-browse
Automate browser tasks like form filling, data extraction, or multi-step web workflows. Use when the user needs to interact with websites that require clicking, typing, or navigation.
browser
Skill "browser" from dhaupin/vant, covering browser automation, when to use, what to do, 1. navigation and 2. interaction.
site-audit
Use when user asks to audit a website, review all pages, check for broken links, capture full site state, or do a comprehensive product review.
automate
Use when user asks to automate browser actions, click buttons, fill forms, navigate pages, run a sequence of browser steps, or build a repeatable workflow.
game-engine
Expert skill for building web-based game engines and games using HTML5, Canvas, WebGL, and JavaScript. Use when asked to create games, build game engines, implement game physics, handle collision detection, set up game loops, manage sprites, add game controls, or work with 2D/3D rendering. Covers techniques for…