Add Bright Data Pro & Rapid Agent specialist

[felixLandlord]

Add Bright Data Rapid & Pro agents — specialized web intelligence via MCP

What does this PR do?

Adds two specialized web-intelligence agents that route all live web work through the Bright Data MCP server:

• Bright Data Rapid Agent — free-tier search, discovery, and markdown scraping (discover, search_engine, scrape_as_markdown, batch variants)
• Bright Data Pro Agent — full Pro MCP toolkit (60+ tools): structured web_data_* extractors, scraping_browser_* automation, and extract

Both agents enforce Bright Data–only web access (no built-in fetch or training-data answers for current web facts), include tier-appropriate retry/fallback paths, and are listed in the Specialized Division table in README.md.

Agent Information

Bright Data Rapid Agent

Agent Name: Bright Data Rapid Agent
Category: specialized
Specialty: Live web research via Bright Data MCP free tier — search, discover, and scrape with citations

Bright Data Pro Agent

Agent Name: Bright Data Pro Agent
Category: specialized
Specialty: Full-spectrum web intelligence via Bright Data MCP Pro — platform extractors, browser automation, and structured field extraction

Motivation

The Agency roster had no dedicated agents for live web research with enforced MCP tooling. General-purpose agents can hallucinate current facts or use inconsistent web access patterns.

These agents fill that gap with a clear tier split:

Need	Agent
Exploratory research, fact-checking, doc lookup on a budget	Rapid — fewer tools, free-tier limits
Amazon/LinkedIn/X pages, JS-heavy sites, interactive flows, structured extraction	Pro — metered but full toolkit

Rapid and Pro share core rules (cite tool output, no non–Bright Data web fallback) but differ in tool scope and cost discipline, so users pick the right agent for the job instead of over-using Pro tools.

Testing

Bright Data Rapid Agent

• Simple research: “I'm evaluating headless browser options for AI agents in 2026. Discover recent comparisons and summarize the top 3 options with pros/cons. Cite every source.” using discover → scrape_as_markdown on top results
• Targeted lookup: "What is the current stable version of Node.js? Use web search, cite the official source URL, and quote the version number from the page you open." via search_engine with inline citations
• Batch comparison: "Run separate searches for: (1) "Bright Data MCP free tier requests", (2) "GitHub Copilot custom agents", (3) "MCP Model Context Protocol". Give one sentence per topic with a source link each." via scrape_batch for a side-by-side summary

Bright Data Pro Agent

• Platform URL: "Using Bright Data MCP only: fetch structured metadata for this public YouTube video URL: Blender 5.2 is AWESOME! — title, channel, and view count if available. Cite the source URL." via matching web_data_* tool
• JS-heavy page: "Using Bright Data MCP only: go to https://news.ycombinator.com/, snapshot the page, click the "More" link if visible, snapshot again, and list the top 5 story titles from the latest snapshot." with scraping_browser_* with snapshot-before-click workflow
• Metered usage: agent batches calls and avoids redundant tool use on multi-step tasks

Shared

• Agent refuses to answer current web facts from training data when MCP is available
• README links resolve to the new agent files

Checklist

• Original — not a near-duplicate (ran scripts/check-agent-originality.sh)
• Follows agent template structure
• Includes personality and voice
• Has concrete code/template examples
• Defines success metrics
• Includes step-by-step workflow
• Proofread and formatted correctly
• Tested in real scenarios