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 Kevin-Liu-01/Agent-Machines --skill computer-usegit clone --depth 1 https://github.com/Kevin-Liu-01/Agent-MachinesWrote 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/kevin-liu-01/agent-machines/computer-use)<a href="https://agentmods.dev/skills/kevin-liu-01/agent-machines/computer-use"><img src="https://agentmods.dev/badge/skills/kevin-liu-01/agent-machines/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/kevin-liu-01/agent-machines/computer-use"><img src="https://agentmods.dev/badge/skills/kevin-liu-01/agent-machines/computer-use.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00039 | $0.00692 |
| Opus 5 | $0.00019 | $0.00346 |
| Sonnet 5 | $0.00008 | $0.00138 |
| Haiku 4.5 | $0.00004 | $0.00069 |
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 9d 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 — 61 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Computer Use (Browser)
Hermes ships with a browser_* toolset backed by Playwright. I can navigate, click, type, scroll, take snapshots, screenshot the page, and read the DOM.
Workflow
- Observe before acting. Always call
browser_snapshotfirst. Don't click on guesses. - Refs are sticky. A snapshot returns refs you can pass to
browser_click(ref=...). Refs invalidate when the DOM mutates -- re-snapshot after navigation, click, type, scroll. - One deliberate step, then verify. After a click that should change state, snapshot again. After a navigate, wait briefly then snapshot. Don't chain four actions and hope.
- Stop after 4 failed attempts. If repeated tries don't progress, surface the obstacle (login wall, captcha, missing element) instead of trying a fifth time.
Common patterns
Login that requires the user. Stop and report. Don't fabricate credentials, don't bypass passkey prompts. The user manually completes the login, then I resume.
Forms with autocomplete that change layout. Type → wait 500ms → snapshot → click the now-correct ref. Never click a ref captured before typing.
Lazy-loaded lists. browser_scroll to bring an element into view. Use scrollIntoView: true rather than scrolling the page repeatedly.
Multi-tab flows. Use browser_tabs(action="list") to enumerate. Switch deliberately.
Coordinate clicks. browser_mouse_click_xy requires coordinates from a fresh viewport screenshot taken immediately before the click for the same tab. Any other tool call invalidates the screenshot.
Anti-patterns
- Polling
browser_snapshotevery 500ms in a loop. The DOM hasn't changed; you're paying for nothing. - Reusing a ref after a click without re-snapshotting.
- Single 30-second
browser_wait. Prefer 2s + snapshot + retry; you proceed as soon as the page is ready. - Trying to interact with iframe content. Hermes can't reach inside iframes. Skip or report.
When I get stuck
Report:
- Current URL.
- What I was trying to reach.
- The blocker observed (with a snapshot or screenshot).
- The most likely next action (manual user step, different selector, different page entry).
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.
- 9d ago First seen · 61 lines · 39 tokens per session scan A 82530061e417
computer-use is a skill published in the GitHub repository Kevin-Liu-01/Agent-Machines (26 stars, last pushed 9d ago), licensed MIT. It adds 39 tokens to every session and 692 once invoked, about $0.0002 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 2026-08-30.
Other skills, from other repositories
webcrawler-deep-crawl
Deep-crawl any website from start URLs, return per-page LLM-ready text/markdown/HTML plus metadata (title, description, author, language, canonical URL, OG) and in-scope outbound links. Use when user mentions deep crawl website, recursive crawl, crawl a whole site, scrape entire website, scrape docs site, scrape…
amazon-search-listing
Amazon search and category listing scraper: extract product listings from any Amazon search results page, keyword search URL, or category browse page and return per-item cards (asin, title, url, image, price, listPrice, stars, reviewCount, badges, isAmazonChoice, isBestSeller, isSponsored, delivery, boughtInPast…
ebay-search-listing
Extracts product listings from any eBay search or category page URL, returning per-item cards (itemNumber, url, title, subtitle, caption, price, priceWithCurrency, currency, wasPrice, bids, shipping, seller, sellerFeedbackCount, sellerPositiveRating, reviewsCount, starRating, image) plus pagination state (currentPage…
facebook-groups-scrape-posts
Scrapes posts from a Facebook group given a group URL, sort order, and desired count — returns structured post metadata including postid, permalink, author, timestamp, body text, images/videos, reaction counts, reaction type breakdown, comment count, and share count. Use when: user wants to scrape/extract Facebook…
facebook-page-profile-posts
Scrapes posts from any public Facebook Page or personal Profile timeline, returning structured data including post text, author info with profile picture, engagement metrics (likes/comments/shares), full reaction breakdown (Like/Love/Wow/Haha/Sad/Angry/Care as both array and flat counts), hashtags and external links…
amazon-bestseller-listing
Amazon Best Sellers listing scraper: extract product cards from any Amazon Best Sellers (zgbs) or /gp/bestsellers/ category page — returns rank (position on chart), asin, title, url, image, imageAlt, price, stars, reviewCount, ratingRaw per item, plus category metadata (categoryName, categoryFullName, categoryUrl) and…