About this skill

Problem

When an Agent must extract data from a specific site or perform browser actions, hard-coded selectors or ad-hoc API probing are fragile. Anti-bot behavior, pagination, session drops, and async rendering can break batch jobs. This is especially hard when the target site lacks public documentation and API parameters depend on page context. browser-act-skill-forge turns a described requirement into a verified, reusable capability package: it explores a stable path first, then produces a deployable SKILL.md plus Python scripts.

How It Works

Requirement confirmation: identify the objective, target site, execution intent, and output directory; if only a business goal is given, search candidate sites and ask the user to choose.
Capability exploration: prefer API endpoints, fall back to DOM when APIs are unavailable, and record reproducible invocation details, enum parameters, and retry rules. If exploration fails, step back to the goal and enumerate alternative means instead of retrying with varied parameters.
Script packaging: convert verified JS snippets into scripts/*.py, expose business parameters via argparse, and validate both normal and error scenarios.
Delivery testing: generate minimal test cases, run them through a Sub-Agent, fix failures, then install or report results.

Boundaries

It only performs actions that the user could manually complete in a logged-in browser and does not bypass authentication or access controls. Intermediate HAR files, temporary records, and extraction results remain on the user's machine. It does not read or write experience notes during generation, and it does not rely on external services to transmit business data. For large-volume extraction, the recommended workflow is to explore and build a Skill first, then execute at scale rather than scraping directly without a proven strategy.

Screenshots

Use Cases

Operations staff periodically extract price-page fields from competitor sites and turn them into reusable scripts.
Analysts with only a business goal search candidate sites, confirm data sources, and generate extraction packages.
Engineers package console filtering, pagination, and export actions into reusable browser capabilities.
Data engineers discover endpoints on undocumented pages and fall back to DOM selectors when APIs are unavailable.

Best For

Data operations staff tracking price and inventory fields on competitor websites
Business analysts repeatedly filtering, paginating, and exporting inside logged-in internal systems
Backend engineers turning ad-hoc web extraction scripts into reusable Agent skills
Scraping engineers discovering endpoints and stabilizing selectors on undocumented sites