web-research

name	web-research
description	How to research URLs, web pages, and topics properly. Fetch first, summarize, give opinion.
agents	["research_agent","job_agent","deep_research_agent"]

Web Research

When Travis sends a URL or asks you to research something, follow this flow:

URL Research Flow

1. Fetch the page using search_web or fetch_page
2. If fetch returns "JavaScript required" or empty content → use browser_navigate to render it
3. Read the full content — don't skim
4. Summarize what you found — key points, not a wall of text
5. Give your opinion if asked ("what do you think")
6. Connect to Travis's context — how does this relate to his projects, skills, goals?

Search Flow

1. Search with specific queries, not generic ones
2. If first search is weak, refine and search again — don't give up after one try
3. Cross-reference multiple sources when possible
4. Always cite where information came from

"Fetch and tell me what you think" Pattern

This means: read the page → summarize → give an honest assessment. Do NOT:

• Create a monitor for the URL
• Apply to a job
• Take a screenshot
• Ask "should I analyze it?"

Just read it, summarize it, say what you think.

Job Posting URLs

When the URL is a job posting (jobs.ashbyhq.com, lever.co, greenhouse.io, careers pages):

1. Fetch and read the full posting
2. Summarize: role, company, requirements, salary, location
3. If Travis asks "what do you think" → assess fit based on his projects and skills in memory
4. If Travis asks to apply → THEN switch to CV tailoring mode
5. Don't auto-apply unless explicitly asked

Failed Fetches

If a page won't load (JS-heavy SPA, blocked, timeout):

• Tell Travis it needs JavaScript rendering
• Offer to open it in the browser tools instead
• Don't return empty results and pretend you analyzed something