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/bigduu/nova/nova-groundingnpx skills add bigduu/Nova --skill nova-groundinggit clone --depth 1 https://github.com/bigduu/NovaWrote 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/bigduu/nova/nova-grounding)<a href="https://agentmods.dev/skills/bigduu/nova/nova-grounding"><img src="https://agentmods.dev/badge/skills/bigduu/nova/nova-grounding.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 | $0.00040 | $0.00799 |
| Opus 5 | $0.00020 | $0.00400 |
| Sonnet 5 | $0.00008 | $0.00160 |
| Haiku 4.5 | $0.00004 | $0.00080 |
Grade A, and why
nova-grounding 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 — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Nova AX-first desktop grounding
Nova controls macOS and Windows. Treat ax_read as the first-class
ax:read capability. It reads Accessibility/UIA directly and does not take
a screenshot. read_ui is a compatibility alias.
Read in this order
- Call
ax_read(window?, mode="all")first for labels, controls, fields, headings, structured text, values, supported actions, and semantic state. It returns an ephemeralsnapshot_id, snapshot-local node IDs, optional bounds, and an explicit coverage/status. - If coverage is absent or partial and the missing information is rendered
text, call focused-window
ocr. OCR returns a grounded center for each recognized line. - Use
screenshot(window=...)orzoom_regiononly when pixels are semantically necessary: layout, icons, colors, images, canvas content, or visual verification.
Do not treat permission_denied as an AX-less application. Grant
Accessibility and retry; a screenshot or OCR call cannot repair the missing
permission.
Act in this order
- Immediately before acting, run a fresh
ax_read, select the exact actionable node, and callax_activate(snapshot_id, node_id). - Nova attempts semantic activation first (
route=axon macOS,route=uiaon Windows, orroute=web_domfor supported browser content). If that is unsupported, Nova may use the node's freshly revalidated center (route=element_center). - For text absent from the semantic tree, click the center returned by OCR
with
left_click(x, y, source="ocr_center"). - Only then click coordinates read from a focused screenshot or zoom
(
source="visual_coordinate").
ax_activate fails closed when its snapshot generation is stale. Every
activation attempt consumes the generation before provider dispatch, including
attempts that return an error because the provider may have partially applied
the action. Run ax_read again after any result and after navigation, refresh,
scrolling, or any other UI change.
click_mark(number=N) remains a compatibility action. Prefer
generation-safe ax_activate. Legacy substring actions (ax_click,
ax_focus, ax_set_value) reject ambiguous matches instead of selecting the
first candidate.
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 First seen · 82 lines · 40 tokens per session scan A 6491b869484b
nova-grounding is a skill published in the GitHub repository bigduu/Nova (17 stars, last pushed 4d ago), licensed MIT. It adds 40 tokens to every session and 799 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
web-navigator
Web 浏览器自动化路由规则。通过 webnavigator 将 Web 场景路由到 Playwright MCP 或 mcp-chrome。.
macos-harness
Control a whole Mac from one persistent Python session with screenshots, PID-targeted input, an animated virtual pointer, targeted Apple Accessibility, Apple Events, Browser Harness CDP, and filesystem access. Use for native, Electron, browser, dialog, file, or cross-app tasks without moving the physical cursor or…
computer-use
Windows 桌面自动化执行规则。使用 screeninteractor 观察与操作原生桌面应用。.
computer-use-macos
Top-level macOS computer-use skill with a bundled standalone runtime that bootstraps itself without any local Claude installation, private native modules, or extracted app assets.
record-workflow
Record an automation workflow by performing actions, then replay it later as a playbook. Capture clicks, typing, navigation into reusable sequences. Use when: "record this", "capture steps", "make repeatable", "save automation", "create playbook", "record and replay", "macro", "record workflow".
fill-form
Fill out web forms with human-like anti-detection typing. Applications, registration, checkout, surveys, contact forms. Use when: "fill out form", "complete application", "register on", "submit form", "fill in details", "checkout", "sign up on".