Browser

What it does

When your agent uses it

• Agent Workflows work in your project

Example prompts

• “/browser”

Workflow steps

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

1. 1Start with the intended action directly (navigate/open/act/evaluate). Do not run status as a pre-check.
2. 2Use snapshot only when refs are required for interaction (click/type/select/drag/scrollIntoView).
3. 3Prefer evaluate for extraction. Return structured data in one comprehensive call when possible.
4. 4Use condition waits by default (loadState/url → selector/text/textGone → fn). Avoid timeMs unless explicitly needed.
5. 5Before clicking potentially off-screen elements, run act.scrollIntoView on the ref first.
6. 6Keep context stable: once targetId is known, pass it in follow-up calls when supported.
7. 7Avoid blind loops: every extra call must have a clear purpose.

What it can do on your machine

Read from SKILL.md and the folder at commit 428f207. 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

Browser loads about 1k tokens when it runs. Until then it costs about 39 tokens; SKILL.md has 304 words of instructions outside code blocks.

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

~39

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

~1k

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 wecode-ai/Wegent at commit 428f207, republished under its Apache-2.0 licence (© wecode-ai). 304 words, ~1,033 tokens.

name

browser

description

Complete real user web tasks end-to-end via browser-tool, navigate, interact, wait for page state, extract results, and provide evidence when needed.

<BROWSER_TOOL_CMD> '<json>'

# Navigate directly
<BROWSER_TOOL_CMD> '{"action":"navigate","url":"https://example.com"}'

# Snapshot only when refs are needed
<BROWSER_TOOL_CMD> '{"action":"snapshot","interactive":true}'

# Act on ref
<BROWSER_TOOL_CMD> '{"action":"act","request":{"kind":"click","ref":"e1"}}'

# Ensure element is visible before click (recommended on long pages)
<BROWSER_TOOL_CMD> '{"action":"act","request":{"kind":"scrollIntoView","ref":"e1"}}'

# Condition wait (preferred over fixed sleep)
<BROWSER_TOOL_CMD> '{"action":"act","request":{"kind":"wait","loadState":"domcontentloaded","timeoutMs":15000}}'

# URL-based wait
<BROWSER_TOOL_CMD> '{"action":"act","request":{"kind":"wait","url":"checkout","timeoutMs":10000}}'

# Run JS in page context via act.evaluate (function or expression)
<BROWSER_TOOL_CMD> '{"action":"act","request":{"kind":"evaluate","fn":"() => ({title: document.title, href: location.href})"}}'

# Run JS against a target element ref via act.evaluate
<BROWSER_TOOL_CMD> '{"action":"act","request":{"kind":"evaluate","ref":"e1","fn":"(el) => ({text: el.textContent?.trim() || \"\"})"}}'

# Close current tab (or pass targetId)
<BROWSER_TOOL_CMD> '{"action":"act","request":{"kind":"close"}}'

# Element screenshot (prefer over full-page when only target proof is needed)
<BROWSER_TOOL_CMD> '{"action":"screenshot","ref":"e1","type":"jpeg"}'

# Comprehensive extraction in one evaluate
<BROWSER_TOOL_CMD> '{"action":"evaluate","expression":"(() => ({title:document.title,url:location.href}))()"}'

© wecode-ai, Apache-2.0. 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 backend/init_data/skills/browser of wecode-ai/Wegent.

Open the folder on GitHubat commit 428f207