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 PeakCode-AI/PeakCode --skill computer-usegit clone --depth 1 https://github.com/PeakCode-AI/PeakCodeWrote 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/peakcode-ai/peakcode/computer-use)<a href="https://agentmods.dev/skills/peakcode-ai/peakcode/computer-use"><img src="https://agentmods.dev/badge/skills/peakcode-ai/peakcode/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/peakcode-ai/peakcode/computer-use"><img src="https://agentmods.dev/badge/skills/peakcode-ai/peakcode/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 | $0.00037 | $0.02381 |
| Opus 5 | $0.00018 | $0.01190 |
| Sonnet 5 | $0.00007 | $0.00476 |
| Haiku 4.5 | $0.00004 | $0.00238 |
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 today.
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 — 185 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Computer Use
The computer tool drives apps on this Mac through a native helper. It is how you do work
that lives in a GUI and nowhere else: a dialog with no CLI, a form in a native app, a
preference toggle, confirming what an app actually shows.
Reach for it last
In order:
- Files and shell commands. Copying a file is
cp, not a Finder drag. - The app's own CLI or config file. Most macOS apps have one, and it is scriptable.
- The
browsertool, if the work is in a web page. It has element refs and a visible pane the user can watch. - Only then
computer.
Skipping to step 4 for something 1–3 could do is slower, more brittle, and it moves the user's cursor around while they are trying to work.
The loop
Observe once, act once, observe again. Never chain GUI actions blind.
action: "list_apps"— what is running. Apps are matched by the name the user would say or by bundle id, so ask for "Finder" even on a system that calls it 访达.action: "get_state"with that app — returns the accessibility tree of the window in front, with a numbered ref for each element, its position and size, and the actions it actually supports.action: "act"with a ref from that tree.- Observe again before concluding anything.
Two other reads, for when the tree is not the right lens:
action: "list_windows"— every on-screen window with its id, app, title and bounds. Use it to find a window that is not the front one (a sheet, a popover, a background window) and to capture it withscreenshot+window_id. Its bounds are already in screen points.action: "displays"— the displays and their scale factors. This is the coordinate space itself: on a second monitor the same pixel is a different point, and this is where that is written down.
get_state also reports a state_id. Pass it back with the ref — a ref is only meaningful
for the observation it came from, and a ref from an older one is refused on purpose. The
element behind it may be a different control by now. Do not retry a refused ref; take a
fresh get_state.
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.
- today First seen · 185 lines · 37 tokens per session scan A ce6ba7230d57
computer-use is a skill published in the GitHub repository PeakCode-AI/PeakCode (134 stars, last pushed today), licensed MIT. It adds 37 tokens to every session and 2,381 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 2026-09-18.
Other skills, from other repositories
orbit-notion
Open Orbit briefing skill — selected by the Orbit pipeline when Notion is the user's only connected connector, or when the user explicitly scopes their daily digest to Notion. Pulls the past 24 hours of document edits, comments, mentions, and database row changes from the user's authenticated Notion connection and…
agentmail
Use your assigned AgentMail inbox to read email tasks, explicitly send or reply, and check delivery. Provided automatically by your inbox assignment.
Cortex
Operate Cortex, the LifeOS memory system — the typed Knowledge Archive (People, Companies, Ideas, Research with typed related: links) plus recall of prior work sessions, ISAs, and conversations. Search, add, harvest, develop, ingest, distill, graph-navigate, recall. USE WHEN cortex, knowledge, knowledge base, search…
pinchtab-mcp
Use this skill when a task requires browser automation through PinchTab's MCP server connected to a remote browser instance. Covers navigation, element interaction, data extraction, form filling, multi-step flows, and session management via MCP tools.
feishu
Work with Feishu or Lark bots, docs, sheets, bitables, approval flows, and OpenAPI/MCP setup without hardcoding credentials.
peekaboo
Capture and automate macOS UI with the Peekaboo CLI.