Skip to content
Skills Library
Personal productivity · Automation

browser-ops-skill

/browser-ops-skill

Triggered when the user asks for page-level interaction with a specific webpage: login, form filling, button clicks, pagination scraping, screenshots, downloading page images, handling CAPTCHAs or anti-bot measures, accessing SPA dynamic content; also used as a fallback when WebFetch returns an empty shell or gets blocked. Typical phrasing: "log me into X then scrape Y", "take a screenshot". **Not for**: factual questions or "what is xx" (→ `WebSearch`), or simply reading static webpage text (try `WebFetch` first).

Repository
openloaf ↗
Licence
AGPL-3.0
Last repository change
14.05.2026

Assessment and security

95/ 100Class: good
Format20 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 14.05.2026, without references to retired models or APIs.

Licence

AGPL-3.0, copyleft: przeróbki udostępniane innym muszą mieć tę samą licencję.

What it does and when to use it

Triggered when the user asks for page-level interaction with a specific webpage: login, form filling, button clicks, pagination scraping, screenshots, downloading page images, handling CAPTCHAs or anti-bot measures, accessing SPA dynamic content; also used as a fallback when WebFetch returns an empty shell or gets blocked. Typical phrasing: "log me into X then scrape Y", "take a screenshot". **Not for**: factual questions or "what is xx" (→ `WebSearch`), or simply reading static webpage text (try `WebFetch` first).

Triggers from the description

log me into X then scrape Ytake a screenshotwhat is xx
Original author description

Triggered when the user asks for page-level interaction with a specific webpage: login, form filling, button clicks, pagination scraping, screenshots, downloading page images, handling CAPTCHAs or anti-bot measures, accessing SPA dynamic content; also used as a fallback when WebFetch returns an empty shell or gets blocked. Typical phrasing: "log me into X then scrape Y", "take a screenshot". **Not for**: factual questions or "what is xx" (→ `WebSearch`), or simply reading static webpage text (try `WebFetch` first).

How it works

  1. Tool Inventory
  2. Core Mental Model
  3. OpenUrl Opening Modes
  4. BrowserSnapshot Details
  5. Three Core Workflows
  6. Selector Selection Strategy
  7. Wait Strategies
  8. Error Diagnosis
  9. CAPTCHAs and Anti-Bot
  10. Iron Rules
Original SKILL.md headings · 112 lines, 809 words
SKILL.md preview
---
name: browser-ops-skill
description: >
  Triggered when the user asks for page-level interaction with a specific webpage: login, form filling, button clicks, pagination scraping, screenshots, downloading page images, handling CAPTCHAs or anti-bot measures, accessing SPA dynamic content; also used as a fallback when WebFetch returns an empty shell or gets blocked. Typical phrasing: "log me into X then scrape Y", "take a screenshot". **Not for**: factual questions or "what is xx" (→ `WebSearch`), or simply reading static webpage text (try `WebFetch` first).
---

# Browser Operations Guide

## Tool Inventory

| Tool | Purpose | Read-only |
|------|---------|-----------|
| `OpenUrl` | Open a webpage (headless / tab / window — three modes) | Yes |
| `BrowserSnapshot` | Get page text + interactive element list + screenshot + rawHtml in one call | Yes |
| `BrowserAct` | Page interaction (click / input / scroll / keypress) | No |
| `BrowserWait` | Wait for page state (load / networkidle / url / text / timeout) | Yes |
| `BrowserDownloadImage` | Download an image from the page to local disk | No |

> **Loading**: All are deferred tools — call `ToolSearch(names: "OpenUrl,BrowserSnapshot,BrowserAct,BrowserWait,BrowserDownloadImage")` to activate their schemas before invoking.

## Core Mental Model

Browser operations = **observe-act-verify** loop. After every action you must snapshot to confirm state, because webpages are stateful — a click may trigger navigation, a popup, or an AJAX load, and you cannot predict the outcome. Blindly chaining actions is the single most common failure mode.

```
OpenUrl → BrowserSnapshot → analyze → BrowserAct → BrowserWait → BrowserSnapshot → ...
```

## OpenUrl Opening Modes

- **`headless`** — First choice for pure automation. No UI; suited for scraping, background agents, and batch operations.
- **`tab` (default)** — Use when the user needs to see/operate the page. Embedded panel.
- **`window`** — Use when the user needs a standalone window for deep interaction.

**Rule of thumb**: user doesn't need to see it → `headless`; user needs to see it → `tab`; user wants a standalone window → `window`.

## BrowserSnapshot Details

A single BrowserSnapshot call returns:
- **Page info**: URL, title, readyState
- **Full text**: body.innerText (truncated at 32KB)
- **Interactive element list**: up to 120 clickable/inputtable elements with their selectors
- **iframe contents**: text and elements from same-origin iframes
- **Screenshot**: captures the full page by default (fullPage), saved to the session asset directory
- **rawHtmlPath**: disk path to the complete outerHTML

**What if text exceeds 32KB?** Use `Read`/`Grep` on `rawHtmlPath` to fetch the full DOM — don't re-snapshot repeatedly.

**Screenshot control**: `fullPage: false` captures only the current viewport.

## Three Core Workflows

### Workflow 1: Information Extraction

`OpenUrl` → `BrowserWait { type: "load" }` → `BrowserSnapshot`

Extract the required information directly from the returned text.

Pagination scraping: `BrowserSnapshot` → `BrowserAct { action: "click-text", text: "Next" }` → `BrowserWait { type: "networkidle" }` → loop. Snapshot on every page.

### Workflow 2: Form Filling and Login

1. `OpenUrl` → `BrowserSnapshot` to inspect form structure and selectors
2. For each field: `BrowserAct { action: "fill", selector: "...", text: "..." }`
3. Submit: `BrowserAct { action: "click-css", selector: "button[type=submit]" }`
4. `BrowserWait { type: "urlIncludes", url: "/success" }` to confirm success
5. `BrowserSnapshot` for final confirmation

**fill vs type**: `fill` atomically clears and inputs — best for forms; `type` appends characters at the current focus — best for search boxes.

**Login caveats**: Cookies persist for the session, so after a single login subsequent requests carry auth automatically. **Do not** snapshot password fields.

### Workflow 3: Screenshots and Image Downloads

- `BrowserSnapshot` — snapshot + full-page screenshot (fullPage by default)
- `BrowserSnapshot { fullPage: false }` — viewport only
- `BrowserDownloadImage { selector: ".product-image" }` — download an image from the page

Screenshots are a debugging superpower: when text doesn't reveal the issue, the screenshot often does at a glance.

## Selector Selection Strategy

Pick selectors from the element list returned by BrowserSnapshot — **don't guess on your own**.

Priority: `#id` > `[data-testid]` > `input[name]` > `.class` > `click-text`

## Wait Strategies

- **`load`**: Use after traditional page navigation.
- **`networkidle`**: First choice for SPA/AJAX pages — safest but slowest.
- **`urlIncludes`**: Wait for redirect after form submission.
- **`textIncludes`**: Wait for asynchronously loaded content.
- **`timeout`**: Last resort.

## Error Diagnosis

**Element not found?** → BrowserSnapshot to confirm page state → may have navigated elsewhere → may be below the viewport (scroll and retry) → may be inside an iframe (cross-iframe not supported)

**Page load timeout?** → BrowserSnapshot to see how far loading got; the content may already be sufficient.

**Action has no effect?** → Page not fully loaded (add BrowserWait) → selector matched the wrong element (check via snapshot) → a popup is blocking (close it first)

**SPA content empty?** → Snapshot again after `BrowserWait { type: "networkidle" }` → if still empty, wait on specific content with `textIncludes`

## CAPTCHAs and Anti-Bot

When you encounter a CAPTCHA, 403/429, or an anti-bot page, **stop immediately and notify the user** — do not retry blindly.

## Iron Rules

1. **BrowserSnapshot after every action to verify state**
2. **Snapshot before acting** — don't guess selectors
3. **Wait for the page to be ready before acting**
4. **When text isn't enough, look at the screenshot**
5. **Don't expose sensitive info** (don't snapshot after filling a password)
6. **Stop immediately on anti-bot encounters**

Versions

RepositoryScoreLicence
This page · openloaf/openloaf95AGPL-3.0
openloaf/openloaf95AGPL-3.0

Database snapshot: 13.11.2026 · package version 88d86644239b (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 90Browser Automation
    This skill should be used when the user needs to interact with web pages, browse websites, take screenshots, fill forms, click elements, extract web content, or perform any browser automation tasks. Trigger phrases include "open webpage", "visit website", "browse", "screenshot…
  • 91browser
    Drives the agent's own browser, or the operator's Chrome for their accounts. Use when WebFetch or WebSearch fails, times out, or returns a 403, an empty page or a JavaScript-only shell; when a page needs clicks, forms, scrolling or sign-in; or when a screenshot or exact rendered…
  • 100docs-fetch-fallbacks
    WebFetch 404/403/timeout/empty-SPA fallback ladder: strip query, raw.githubusercontent, `gh api`, docs-source-at-a-pinned-tag, context7/WebSearch. Use when a doc or GitHub URL fetch fails or returns a page with no content.
  • 96web-form-filling
    Use when filling SPA web forms through browser_exec.
  • 100fetch
    Default, free way to read a webpage or any URL's actual content: fetch the full page as clean markdown, HTML, or structured JSON, including JavaScript-rendered pages, for up to 10 URLs in parallel. Use to read or summarize a webpage or article, extract data from a website, web…
  • 95macos-control-skill
    Triggered when the user asks the AI to directly operate their macOS desktop: open an app, click a button, fill a form inside a native window, read what's currently on screen, automate a local GUI flow. Typical phrasing: "in Finder, please…", "click X in the Y app", "what's on my…

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.