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 EZotoff/ez-omo-config --skill computer-usegit clone --depth 1 https://github.com/EZotoff/ez-omo-configWrote 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/ezotoff/ez-omo-config/computer-use)<a href="https://agentmods.dev/skills/ezotoff/ez-omo-config/computer-use"><img src="https://agentmods.dev/badge/skills/ezotoff/ez-omo-config/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/ezotoff/ez-omo-config/computer-use"><img src="https://agentmods.dev/badge/skills/ezotoff/ez-omo-config/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 | $0.00082 | $0.01824 |
| Opus 5 | $0.00041 | $0.00912 |
| Sonnet 5 | $0.00016 | $0.00365 |
| Haiku 4.5 | $0.00008 | $0.00182 |
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 4d 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 — 131 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Computer Use (cua-driver)
You control the local Linux X11 desktop through the skill_mcp tool with
mcp_name: "cua". One tool per call. Keep image-producing calls on the CLI
path (explained below); non-image example:
skill_mcp(mcp_name="cua", tool_name="get_window_state", arguments={"pid": ..., "window_id": ..., "include_screenshot": false})
Infrastructure facts (already deployed, do not change):
- Daemon: systemd user service
cua-driver.service, pinned 0.20.0, overlay disabled (it froze GNOME Shell on this dual-head — never re-enable), telemetry off. - The daemon is shared and always-on; your calls are cheap and isolated per session label.
Decision ladder — when to use what
| Task shape | Tool |
|---|---|
| Target app has a CLI, config file, or API | bash — never this skill |
| Web — ANY web, logged in or not | agent-browser (persistent named session for logins; one-time manual login) — never this skill |
| Native GUI app (GNOME/Qt/Electron) | this skill, AT-SPI element rung first |
| OS chrome (dialogs, polkit, keyring, settings) | this skill, pixel rung |
Action method (strict order)
- Element rung (preferred).
list_windows→ find target →get_window_state(returnselements[]withelement_token+ the app tree) → act byelement_token(click,set_value,type_text...). No coordinate math needed. Re-read the tree after any UI change — tokens go stale. Cross-check the tree against pixels when it looks wrong: Electron apps echo-confirm and virtualized lists report bogus geometry (the tree sometimes lies). Backgroundtype_text: AT-SPI EditableText lands in UNFOCUSED GTK4/Qt6 editables — no focus needed; terminals take a focus-free pty route. Non-editable focused widgets (canvas, spreadsheet cell) need the foreground rung. - Pixel rung (fallback). The screenshot from
get_window_stateis in window-local pixels — pixel actions take x,y relative to that window's screenshot, NOT the desktop. Ground coordinates withlook_aton the saved image when unsure. Click centers of elements; on miss, re-shoot, adjust, retry. Apply the ~47px header offset before clicking (see Failure modes).
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.
- 4d ago First seen · 131 lines · 82 tokens per session scan A f7c4051c0396
computer-use is a skill published in the GitHub repository EZotoff/ez-omo-config (5 stars, last pushed 3d ago), licensed MIT. It adds 82 tokens to every session and 1,824 once invoked, about $0.0004 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 2026-09-04.
Other skills, from other repositories
minimal-code-discipline
YAGNI-ladder coding discipline for writing changes — existence-first, reuse before writing, dependency ladder, shortest correct diff, with non-negotiables that must never be minimized away.
armada-verification
Mandatory pre-completion verification and evidence checklist before reporting done.
armada-resume
Resume a killed or interrupted armada session. Use on session start or when nextAction is non-empty. Triggers on: resume, reconcile, killed session, /armada-resume.
autopilot
Full autonomous execution from idea to working code.
graph
Deterministic orchestration graph runtime - declarative DAG pipelines with journal-based crash recovery.
ai-slop-cleaner
Clean AI-generated code slop with a regression-safe, deletion-first workflow and optional reviewer-only mode.