← All use cases

中文

Operations & admin · Claude

Research your building's violations and draft the complaints

Idea + starter prompt, not run by us yet

For: Renters who suspect their landlord is breaking housing rules

A poster had a Claude agent play tenant adviser. It pulled New York City's violation records and building history, read the housing code, tied each issue to a section and drafted 14 complaints, filed on a spread-out schedule and tracked in a small database. The poster says the city fined the landlord nineteen days later and blocked the rent rise.

Steps

  1. List the problems in your unit with dates and photos.
  2. Find your city's public building-violation database and ask Claude to pull the history for your address.
  3. Ask it to match each problem to the exact section of the local housing code and say if it is unsure.
  4. Review every complaint and send them yourself. The source filed them spaced out rather than all at once.
  5. Keep a log of complaint numbers and replies. Speak to a tenant advice service before any legal step.

Starter prompt

Starter prompt. Change the [brackets] first.

Trial first: run the whole job on a made-up address, [made-up address], in New York City, with two invented problems. Save the drafts in [trial-complaints]. Use my real address only after I say yes.

I rent a flat at [address] in New York City. Problems I have found: [list with dates, and I have photos]. Search the city's public building-violation records for this address and summarise what is already recorded. Then find the part of the housing code that covers each problem and draft a short complaint for each one. Show me the drafts and section numbers, and say where you are unsure.

Stop and wait for my yes before anything is filed. A draft is a pause, not permission. I file every complaint myself, spaced out. You are not a lawyer, so give no legal advice. Do not ask for my ID or passwords.

Original post: X · @Argona0x · found via agentusecases.ai. Written up in our own words. We haven't run it, so test it on made-up data first.