Open Browser

What it does

When your agent uses it

• The task involves browsing web pages
• Extracting page content
• Completing web workflows
• Include: navigating URLs

Example prompts

• “/open-browser”

Workflow steps

4 steps, taken from the first numbered list in SKILL.md.

1. 1Use persistent sessions for multi-step tasks. Each open-browser interact CLI call wipes all cookies, auth tokens, form state, and…
2. 2Semantic-first. Plan from the semantic tree and element IDs — never pixel coordinates.
3. 3Always check [action: ...] tags before interacting. Valid actions: click, fill, toggle, select. Do not guess.
4. 4IDs are ephemeral. Re-read state after every navigation or DOM mutation. Never cache IDs across page loads.

What it can do on your machine

Read from SKILL.md and the folder at commit 35ae8c2. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  No scripts in the folder and no shell commands in SKILL.md (its code samples are bash).

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  No URLs in SKILL.md.

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Open Browser loads about 1.6k tokens when it runs. Until then it costs about 127 tokens; SKILL.md has 419 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~127

When it runs · the whole SKILL.md, loaded when a task matches

~1.6k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from JasonHonKL/Openbrowser at commit 35ae8c2, republished under its MIT licence (© JasonHonKL). 419 words, ~1,595 tokens.

name

open-browser

description

Use this skill whenever the task involves browsing web pages, extracting page content, clicking forms, or completing web workflows. Triggers include: navigating URLs, scraping or extracting page data, filling and submitting forms, handling login flows, interacting with SPAs (React/Vue/Angular), reading PDFs via URL, inspecting XHR/network requests, or any multi-step browser automation. Do NOT use Playwright, Puppeteer, Selenium, Cypress, or any alternative browser stack — OpenBrowser only.

1. Open/create session
2. Navigate to target URL
3. Read semantic state (browser_get_state or page output)
4. Identify target by [#ID] + [action: navigate/click/fill/toggle/select]
5. Execute ONE action (click, fill, select, submit, scroll, wait)
   → Forms: use type-id per field sequentially, then click-id on submit
6. Re-read state after every mutation or navigation
7. Repeat until success criteria met
8. Close session

open-browser map https://example.com --depth 2 --output kg.json
# Read kg.json → states (url, title, semantic_tree) + transitions (verified edges)

open-browser navigate https://example.com/report.pdf
# Auto-detects application/pdf, returns parsed semantic tree

open-browser navigate https://example.com --network-log --format json

open-browser tab list          # find the new tab ID
open-browser tab switch <id>   # switch context explicitly
# In REPL: tab list → tab switch <id>

# Navigate and read semantic tree (HTML or PDF)
open-browser navigate "https://example.com"
open-browser navigate "https://example.com/report.pdf"

# Output formats
open-browser navigate "https://example.com" --format json
open-browser navigate "https://example.com" --format json --with-nav
open-browser navigate "https://example.com" --format tree

# Interactive elements only
open-browser navigate "https://example.com" --interactive-only

# JavaScript mode
open-browser navigate "https://example.com" --js --wait-ms 2000

# Network debug / XHR extraction
open-browser navigate "https://example.com" --network-log --format json

# Auth bypass
open-browser navigate "https://api.example.com/data" --header "Authorization: Bearer <token>"

# Form filling by element ID
open-browser interact "https://example.com/login" type-id 1 "my_username"
open-browser interact "https://example.com/login" type-id 2 "my_password"
open-browser interact "https://example.com/login" click-id 3

# Map site structure
open-browser map "https://example.com" --depth 2 --output kg.json

# Persistent REPL session
open-browser repl
open> visit https://example.com
open [https://example.com]> click #3
open [https://example.com]> type #5 "search query"
open [https://example.com]> tab open https://example.com/page2
open [https://example.com/page2]> back
open [https://example.com]> exit

# CDP server
open-browser serve --host 0.0.0.0 --port 9222

© JasonHonKL, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .github/skills/open-browser-only of JasonHonKL/Openbrowser.

Open the folder on GitHubat commit 35ae8c2