headsvk

Claude Flat Search — Development skill for Claude Code

Development community

Searches Rightmove, Zoopla, OnTheMarket and OpenRent in a real browser and reads the listing prose for what their filters cannot express: floor area, not a basement, a lift above the third floor, air.

How to install Claude Flat Search

This entry records only its repository, not the path inside it, so there is no exact command to give. Open headsvk/claude-flat-search and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Claude Flat Search does

Searches Rightmove, Zoopla, OnTheMarket and OpenRent in a real browser and reads the listing prose for what their filters cannot express: floor area, not a basement, a lift above the third floor, air conditioning in the flat rather than the residents' gym. Claude judges the ambiguous cases and must quote the page; unbacked claims are discarded.

Alternatives in Development

  • Portaljs Add Chart — Add a chart (line, bar, area, pie, or scatter) to a dataset's showcase in a PortalJS portal 2.3k ★
  • Pinloop CLI — Command-line job search tool for coding agents 317 ★
  • Simplify 스킬 — 유형: 번들 스킬 (Claude Code 바이너리에 컴파일됨) 호출 방식: /simplify 또는 /simplify focus on [area] /simplify 명령은 최근 변경된 파일에 대해 자 149 ★

README

UK flat search, read by Claude

A rental search that runs itself every morning and hands you a short list of only what is new, ranked, with the things the portals will not let you filter on already read out of the listing text.

It exists because the filters on Rightmove, Zoopla, OnTheMarket and OpenRent cannot express what most people actually want. You can filter on bedrooms and price. You cannot filter on *floor area over 800 sq ft*, *not a basement*, *has a lift if it is above the third floor*, or *air conditioning in the flat rather than in the residents' gym*. Those live in the prose, so this reads the prose.

Four portals, one browser each, run concurrently. How many listings that comes to is entirely a function of your criteria — a wide net over a city runs to several hundred tracked at a time, a narrow one to a few dozen. Either way a morning run turns up a handful of genuinely new ones, which is the number that matters.

All four portals are national, and nothing in the filtering knows what a London postcode is, so this works anywhere in the UK. The worked example is London because that is where it was built and measured; [Outside London](#outside-london) is the two things you change.

**You do not have to set this up by hand.** Everything below — installing it, writing the config, working out the search URLs, running it on a schedule — is work you can hand to a coding agent. See [Let an agent set it up](#let-an-agent-set-it-up).

How it works

search  ->  audit  ->  plan  ->  details  ->  judge  ->  commit  ->  render
Stage What it does
search Pages each portal's search results in a real browser. Cheap fields only.
audit Per-portal field coverage. Refuses to certify a portal it has not sampled.
plan Dedupes against what is already tracked, applies the hard filters, emits a fetch queue.
details Fetches each listing page and extracts the description, amenities and floor area.
judge Decide