Company history: how we keep the page

Aquiba Benarroch · 3 October 2026 · For Allan and William Barnes

Agents have been excellent at finding pages and unreliable at keeping them. A subagent found a page, sent a summary, and the summary became the source. An image that held the fact could be missed. Will's Perimeter Solutions pieces needed every quote checked again. The fix is to save the page first, the way we save a filing, and let an agent read the file.

1. What a terminal skill does

flowchart LR
  q["Question"]
  find["Skill finds a page"]
  sum["Summary comes back"]
  src["Summary is treated as the source"]
  q --> find --> sum --> src
  

2. The script

flowchart LR
  q["Question: founding, ownership, milestones"]
  fetch["Firecrawl fetches the page"]
  file["File saved"]
  row["A row for every URL found, including ones not opened"]
  q --> fetch --> file --> row
  

3. The agent sits on the files

flowchart LR
  files["Saved files"]
  agent["Agent queries them"]
  quote["Quote points at the file"]
  files --> agent --> quote
  

Numbers read 3 October 2026 from ~/Code/company-history-capture/captures/costco/20261003T093324/manifest.jsonl. Allan's note is in InvestOS plan / priorities.

Appendix

Fetched (12). Under each link: history yes or no, and the one quote we kept.

Schema we asked Firecrawl to fill

On each opened page, one JSON object. Required field: page_is_about_company_history. If that is false, the arrays stay empty. Other fields: founded_year, founders, ownership_changes (year, event, parties, quote), milestones (year, event, quote), name_or_legal_changes. The prompt asked only for founding, founders, headquarters origin, ownership changes, name or legal-entity changes, and dated operating milestones. Every fact needs a quote from that page.

Not fetched (26). Same manifest. Not opened.