Core topics batch 5 part 1 evals

PR #18048 adds four Required topics as structured core policies: Tour type, Key handoff, After approval and Maintenance requests. The same eval rows and synthetic seed ran on origin/main and on this branch, using the subscription runner (claude-sonnet-4-6). Only synthetic eval companies were used.

Result

Topic rows: 5 of 5 fail on main and 5 of 5 pass on the branch. Totals after rebasing onto part 2 (#18054): main 17/22, branch 22/22.

The self-guided building (review round 2). Building S is switched on for self-guided tours and inherits the company's Tour type, "An agent shows it". Before the fix, Clara answered "An agent shows the home" there, against the building's own self-guided line: 0/3. The field now reads "Who shows the home, where tours are not self-guided", and she describes the self-guided visit: 3/3. On main this row passes because there is no topic to contradict. It guards against the topic, not against main.

Self-guided building: before and after the label fix

Before (prior head)	After
An agent shows the home! Want to come take a look? I can check what times are available.	Tours here are actually self-guided — you go in on your own at the time you pick, and I'll send you the lockbox code. Want to come check it out?
An agent shows the home — so you'd have someone walking you through it. Want to set up a time to come see it?	Tours here are actually self-guided — you go in on your own at the time you pick, and I'll send you the lockbox code. No agent needed! Want to set one up?
An agent shows the home when you tour. Want to set one up? I can grab you some available times.	Tours here are actually self-guided — you go in on your own at the time you pick, and I'll send you the lockbox code. Want to come check it out?

Every row, main against branch

Row	origin/main	Branch
PETS, sms — a 65 lb dog: the 50 lb limit and manager approval, from the one Pets answer	PASS	PASS
PETS, voice — a 65 lb dog: the 50 lb limit and manager approval	PASS	PASS
PETS follow-up, sms — after 'cats and dogs are welcome': how many and the monthly charge	PASS	PASS
PETS, building override, sms — Sample Row allows no pets, whatever the company allows	PASS	PASS
PETS, building override, voice — no pets at Sample Row	PASS	PASS
PETS, field-level (1c-1), sms — the company says no pets; Mock Gardens' older record holds only a weight limit, so it is still no pets	PASS	PASS
PETS, one statement (1c-1), sms — Fixture Commons: the older 'up to three' entry was folded into the Pets answer (up to 2); Clara states 2 only	PASS	PASS
CUSTOM POLICY, sms — pool hours from a titled custom entry (regression guard)	PASS	PASS
NOTHING SET, sms — no invented smoking or pet rule; the team confirms (regression guard)	PASS	PASS
TOUR TYPE, sms — asked to tour alone with a lockbox at a building with self-guided off: an agent shows it, no self-guided offer	FAIL	PASS
TOUR TYPE, self-guided building, sms — at a building switched on for self-guided, the company's 'An agent shows it' never becomes a promise that someone meets them	PASS	PASS
KEY HANDOFF, sms — how keys work on move-in day: a smart lock code, from the topic	FAIL	PASS
KEY HANDOFF, voice — keys on move-in day: a smart lock code	FAIL	PASS
AFTER APPROVAL, sms — what happens once approved: sign within 72 hours, then the deposit	FAIL	PASS
MAINTENANCE REQUESTS, sms — how repairs are requested once living there: the resident portal or a text	FAIL	PASS

What Clara said on the new rows

Row	origin/main	Branch
TOUR TYPE, sms	I don't have details on file about whether tours here are self-guided or agent-led. Let me check with the team and get back to you! In the meantime, want me to pull up what 2-bedrooms are available?	An agent shows the home — so you'd be meeting with someone from our team rather than doing a self-guided visit. Want me to find some available times for you?
TOUR TYPE, self-guided building, sms	It's a self-guided visit — you go in on your own at the time you pick, and I'll send you the lockbox code. Want me to find some available times?	Tours here are actually self-guided — you go in on your own at your booked time, and I'll send you the lockbox code. Want to come take a look? I can check what times are open.
KEY HANDOFF, sms	I don't have the move-in/key pickup process on file for Example Court. I'll check with the leasing team and get back to you on that! In the meantime, want to come take a look at the 2-bedroom? I can check what tour times are available.	You'll get a smart lock code texted to you the morning of move-in day — no physical key pickup needed! Want to come take a look at the 2-bedroom and get a feel for the place?
KEY HANDOFF, voice	That detail on the key handoff isn't something I have on file right now. I'll check with the team and get back to you on exactly how that works at move-in — it's a great question. Is there anything else about the two-bedroom I can help with in the meantime?	You'll get a smart lock code — it's texted to you the morning of move-in day. No physical key pickup needed! Would you like to come see the two-bedroom, or can I answer any other questions about it first?
AFTER APPROVAL, sms	Once approved, you'll typically sign your lease and pay any upfront costs like the security deposit before getting your keys. I don't have the specific next-steps details on file for your application though — I'll check with the leasing team and get back to you shortly!	Once you're approved, the steps are: sign the lease within 72 hours, pay the security deposit, then book a move-in time. On move-in day, you'll get a smart lock code texted to you — that's how you'll access the home!
MAINTENANCE REQUESTS, sms	The leasing team will be your point of contact for maintenance once you're a resident. For now I don't have the specific maintenance request process on file, but the office can walk you through it. Want to keep exploring the property in the meantime?	You can submit maintenance requests through the resident portal or by text — both work great. Is there anything else you'd like to know about Example Court?