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/wende/macbeth/safarinpx skills add wende/macbeth --skill safarigit clone --depth 1 https://github.com/wende/macbethWrote 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/wende/macbeth/safari)<a href="https://agentmods.dev/skills/wende/macbeth/safari"><img src="https://agentmods.dev/badge/skills/wende/macbeth/safari.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 | $0.00018 | $0.01698 |
| Opus 5 | $0.00009 | $0.00849 |
| Sonnet 5 | $0.00004 | $0.00340 |
| Haiku 4.5 | $0.00002 | $0.00170 |
Grade A, and why
safari scanned grade A with 1 finding 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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
import { execSync } from "node:child_process"; How it starts
The opening of the file, as written. The whole thing — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Safari Automation
Connect
Always connect with: connect_app({ name: "Safari" })
Key Concepts
Safari's AX tree has two main areas:
- Chrome (toolbar, tabs, menus) — native macOS UI elements
- Web content — a
web_areaelement containing the page's DOM-like accessibility tree (headings, links, groups, text, lists, tables, etc.)
Navigation
Navigate to a URL — fill on the address bar sets the AX value but does NOT trigger navigation. Use open via shell instead:
open -a Safari "https://example.com"
Or with the macbeth SDK:
import { execSync } from "node:child_process";
execSync('open -a Safari "https://example.com"');
Read current URL:
{ "query": [
{ "role": "window" },
{ "role": "toolbar" },
{ "role": "text_field", "identifier": "WEB_BROWSER_ADDRESS_AND_SEARCH_FIELD" }
]}
The value field contains the current URL.
Back / Forward: Click the toolbar buttons:
{ "query": [{ "role": "window" }, { "role": "toolbar" }, { "role": "button", "identifier": "BackButton" }] }
{ "query": [{ "role": "window" }, { "role": "toolbar" }, { "role": "button", "identifier": "ForwardButton" }] }
Reload: select_menu_item({ app: "Safari", menuPath: ["View", "Reload Page"] }) or click ReloadButton
Common Queries
Toolbar Buttons
| Element | Query |
|---|---|
| Address bar | [{ "role": "window" }, { "role": "toolbar" }, { "role": "text_field", "identifier": "WEB_BROWSER_ADDRESS_AND_SEARCH_FIELD" }] |
| Back | [{ "role": "window" }, { "role": "toolbar" }, { "role": "button", "identifier": "BackButton" }] |
| Forward | [{ "role": "window" }, { "role": "toolbar" }, { "role": "button", "identifier": "ForwardButton" }] |
| Reload | [{ "role": "window" }, { "role": "toolbar" }, { "role": "button", "identifier": "ReloadButton" }] |
| New Tab | [{ "role": "window" }, { "role": "toolbar" }, { "role": "button", "identifier": "NewTabButton" }] |
| Share | [{ "role": "window" }, { "role": "toolbar" }, { "role": "button", "identifier": "ShareButton" }] |
| Sidebar | [{ "role": "window" }, { "role": "toolbar" }, { "role": "button", "identifier": "SidebarButton" }] |
| Tab Overview | [{ "role": "window" }, { "role": "toolbar" }, { "role": "button", "identifier": "TabOverviewButton" }] |
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 134 lines · 18 tokens per session scan A 3bc112f8fd74
safari is a skill published in the GitHub repository wende/macbeth (3 stars, last pushed 9d ago), licensed MIT. It adds 18 tokens to every session and 1,698 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
mirroir-onboard
Onboard a consumer web app to mirroir's .mirroir/ dotfile by EXPLORING the running app (chrome-devtools-mcp) — derive real selectors from the accessibility tree, exercise each surface's primary action, emit the .mirroir/ tree, and validate by LIVE REPLAY with a self-heal loop. Reject shallow "page renders" coverage.
@elad12390/web-research-assistant
Comprehensive MCP server for web research with 14 tools: SearXNG + Exa AI search, URL crawling, stealth scraping (Cloudflare bypass), package registry lookup (npm/PyPI/crates), GitHub stats, changelog fetching, tech comparison, error translation, API docs discovery, stock image search, and service health. Trigger…
verify
Drive the running ClawStash app in a real browser to verify UI/UX changes end-to-end. Use after nontrivial frontend changes, before committing — tests and tsc alone have missed lifecycle/timing bugs in this repo before (see MEMORY.md.
openclaw-web-scraper
Stealth web scraping MCP server with 5 tools for structured data extraction. Use when: (1) 'scrape this URL' or 'extract data from this site', (2) 'get all prices from this page' or 'extract structured data', (3) 'take a screenshot of this page', (4) 'click this button and get the result' or 'interact with a page'…
browser-testing-with-devtools
Verify browser behavior with runtime evidence from DOM, console, network, and performance signals.
opencli-sitemap-author
Use when creating or maintaining OpenCLI site sitemaps: agent-facing navigation, page-state, action, workflow, API-reference, pitfall, and fallback knowledge for a website. Use after browser exploration discovers durable site context, when a sitemap is stale, or when promoting local site knowledge into the repo.