Dashboard/PRISM (Second-Brain)/wiki/entities/scout.md
PRISM (Second-Brain)

wiki/entities/scout.md

Scout

Overview

Scout is PRISM's evidence-acquisition engine. It acquires the target company's OWN first-party data: company info, executives, careers and jobs, and investor relations. It scrapes their site, including JS-rendered pages, sitemaps, and PDFs.

Key Capabilities

  • First-party data acquisition with citations. Empty citations mean not grounded.
  • Handles JS-rendered pages and PDF extraction that a basic fetch cannot.
  • Co-hosted on the VPS; PRISM calls it over HTTP on loopback (8421).

How PRISM uses Scout

PRISM uses Scout to acquire the target's own data. PRISM never reimplements crawling; it calls Scout and interprets the result. Scout acquires and cites; PRISM interprets. This is the acquisition plane of the right-tool-for-each-job doctrine.

Caveat: Scout can return success with empty markdown on some sites (Squarespace, PDFs). Check raw_html.

  • prism uses Scout for first-party target data.
  • Companion sources: gemini-grounded-search (open web), detect-search (tech), yfinance (financials), chrome-browser (live search test).