Browser

Dia Ask Public

Try it

Prompt The Browser Company's Dia (a macOS browser whose assistant is an AI agent) from the command line and get its answer back as an exact text file. Use to read/research logged-in or JS-heavy pages in a real browser session. macOS + Accessibility only; unofficial UI automation.

What it does

Prompt The Browser Company's Dia (a macOS browser whose assistant is an AI agent) from the command line and get its answer back as an exact text file. Use to read/research logged-in or JS-heavy pages in a real browser session. macOS + Accessibility only; unofficial UI automation.

The skill document

dia-ask — drive Dia's assistant from the CLI

Delegate a prompt to the AI assistant inside a locally-running Dia browser and read its answer as a file. Dia renders logged-in, JavaScript-heavy pages in your real session and writes its answer to disk, so you get exact text — any length, any format — without screen OCR.

Dia is eyes, not hands: reach for it to read / research what's behind a login or heavy JS; use other tools to act (click, submit).

Built by AgentNeo — AI agents with operational rigor, and the open-source safety patterns to run them.

Requirements

  • macOS, with Dia installed (runs on your own Dia + subscription — no hosted service).
  • Accessibility permission for the process running node (System Settings → Privacy & Security → Accessibility).
  • Node.js ≥ 18. No dependencies.

Usage

node dia-ask-v2.js "" [--format md|txt|json|csv] [--timeout 300] [--no-fallback] [--debug]

Prints one line to stdout: the absolute path of the file Dia wrote. Read that file.

node dia-ask-v2.js "summarize the open tab and list every external link" --format md
node dia-ask-v2.js "extract every row of the pricing table on the page" --format csv --timeout 600
  • dia-ask-v2.js — default, focus-safe (injects into Dia's existing window via CGEventPostToPid; never steals focus).
  • dia-ask.js — v1 fallback (opens a fresh window, briefly takes focus). v2 falls back to it automatically; --no-fallback disables.

Caveats

Unofficial UI automation: it depends on Dia's window/Accessibility internals and can break on any Dia update. No official API. macOS only. A reference implementation, not a hardened product. See README.md.

About

Built and maintained by AgentNeo — we build AI agents with operational rigor and open-source the safety patterns needed to run them in the real world.

Contact: gk@agneo.app

Related skills

Generate precise static or interactive geometry diagrams for geometry problems using GeoGebra's Execute text script. Use when Codex needs to create a constru...

15 installs

Browser automation CLI for AI agents. NEVER run browser-act commands directly via Bash — always invoke this skill first. Use browser-act when a user mentions...

9 installs1 stars

Use this skill when an agent needs to read or navigate websites without a browser: fetch a page as clean Markdown, get a JSON accessibility snapshot of inter...

9 installs

Use when AI agents need to read, inspect, operate, navigate, debug, or verify browser state through the omnibot CLI and connected Chromium extension.

1 stars

Browser Use (browser-use.com). Use this skill for ANY Browser Use request — reading, creating, and updating data. Whenever a task involves Browser Use, use this skill instead of calling the API directly.

2 installs

Anchor Browser (anchorbrowser.io). Use this skill for ANY Anchor Browser request — reading, creating, and updating data. Whenever a task involves Anchor Brow...