Add Petfinder mirror site

[Sun-sunshine06]

Summary

• add a Petfinder mirror site with browse, detail, search, tools, guide, login, account, and save workflows
• register the site as port 40015 and add build-time seed generation in Dockerfile
• include 10 benchmark tasks for the mirror

Verification

• sites\petfinder\venv\Scripts\python.exe -m py_compile sites\petfinder\app.py
• python -m py_compile control_server.py
• parsed sites/petfinder/tasks.jsonl: 10 tasks, all using http://localhost:40015/
• Flask test client smoke: /, /listings, /tools, /guides, /search?q=New York, /login all returned 200
• login/save smoke with alice.j@test.com / TestPass123! passed
• seed byte identity check passed: 1e6d85a93f153b6a5916a8a968cee994

[Susan571]

Reviewer checklist (WebHarbor Review Guide)

Reviewer note: Source-level review (checked out pull/43/head; read app/templates/seed/tasks). The mechanical Docker pipeline (build.sh, container run, curl 200, md5 reset) was not run in my environment — please pair with a Docker-capable mechanical pass. The claimed reset md5 1e6d85a9… is unverified.

Visual fidelity: ❌ FAIL

• This is a generic CONFIG-driven template, not a Petfinder mirror. base.html is ~28 lines; index.html/listing_detail.html are 1–5 lines rendering "Featured {{config.entity}}s" cards — no resemblance to petfinder.com layout/IA.
• All imagery is the /art/<slug>.svg placeholder (two circles + brand initials "PE"). No real pet photos → fails the "no placeholders/colored rectangles" rule.

Functional depth: ⚠️ Weak pass

• Working: token-scored search, category/region/price filters, detail pages, login, save-pet, account preferences (auth + 1 CRUD + 1 state).
• Domain model is wrong: adoptable pets are given a price ($20–$200) and a score (xx/100). Petfinder is an adoption site — pets aren't priced or scored. Summaries are formulaic filler ("Milo … is a dogs result for Petfinder users in New York, rated 96/100"), not scraped catalog data.
• No register route (login only).

Task quality: ❌ FAIL (10 tasks; guide wants 15–20)

• Petfinder--0 "open the highest-scoring pet, report its score" — score is a fabricated field and the homepage is already ordered by score → answer leak / trivial.
• Petfinder--9 "Search for Distance and report one result" — "Distance" is a filter label, not content → broken/nonsensical.
• Petfinder--4 leaks the password (TestPass123!) in the prompt.
• All tasks are single-step (search→open / filter→open): no near-miss distractors, no ≥5-step reasoning.

Verdict: Request changes

Please: (1) mirror real Petfinder structure + use real pet photos; (2) adopt an adoption-appropriate data model (breed/age/size/shelter/location instead of price/score); (3) provide 15–20 leak-free, multi-step tasks (≥3–5 needing ≥5 steps) with distractors; (4) fix the broken --9 task; (5) remove credential leaks from task prompts.

[jackjin1997]

Hi @Sun-sunshine06 — claiming this PR for review.

I will take over the reviewer-owned continuation while preserving your original commit and attribution. The review will start from @Susan571 source-level findings, then cover the missing mechanical Docker pass, Petfinder source/UI fidelity, functional and reset behavior, real UI execution for every accepted task, and a deterministic verifier for each accepted task.

The current branch is conflicting with main and still reflects the older 10-task/build-time-seed shape. I will re-integrate it on a separate Review: branch, record issues before changing them, keep only tasks that meet the current contract rather than padding the count, and link the Draft review PR back here. I will also separate the GitHub code candidate from any required Hugging Face asset delivery.

Please let me know if someone is already actively taking this over so we can avoid duplicated work. The maintainers remain responsible for final approval and merge.

[jackjin1997]

Draft reviewer continuation is now open: #113. The corresponding reviewed asset bundle is in Hugging Face PR https://huggingface.co/datasets/ChilleD/WebHarbor/discussions/80. Both remain unmerged; independent blind review and final reconciliation are still pending.