Session Handoff — Base44 stuck-app conversions: decoded the "Migrated from Lovable" shells + wrote Fable build/fix prompts

WHERE WE JUST WERE — verbatim (MANDATORY, always first)

Final substantive message (LeadScout fix prompt delivery):

> Grounded — and the entity schemas + page list pin all three of your symptoms exactly, no guessing needed: > > 1. Can't open a campaign / see its results → the app has pages for Home, Campaigns, Leads, Settings — but no campaign-detail page exists at all. You can launch a scan, but there's no screen that shows that one campaign's results. > 2. Dashboard cards don't click → same missing destination; Home shows the counts but nothing links anywhere. > 3. Zip codes only → the Campaign data model literally requires zip_codes (+ a per_zip_limit), so the form forces zips. City/county isn't supported anywhere — but Outscraper's map search takes free-text ("niche in city, ST"), so it's fixable. > > Real app, real data (your 2 campaigns + 133 leads) — so the prompt tells Fable to preserve it. Saved to PROJECTS/LeadScout Studio/base44-leadscout-original-fix-fable-prompt.md. > > [Full LeadScout Fable prompt printed as a ``markdown copy block — identical to the saved file.] > > That's the LeadScout fix ready to fire alongside your other sessions. > > 🧠 One more time, plainly: this session is at ~262K tokens, past your 250K freshness line. Good moment to run /session-handoff` — you've now got 4 Fable prompts out (WallCraft, Momentum, Sendji, LeadScout) plus the Big Worm one still queued, and a handoff will keep the next stretch sharp.

Prior substantive message (Sendji prompt delivery — key diagnosis, verbatim):

> Confirmed — and Sendji is a different animal from the other two. The decisive finding: > App.jsx has zero routes wired — the <Routes> block is still Base44's empty scaffold ({/ Add your page Route elements here /}), so every URL renders PageNotFound. Nothing is reachable. And the core product screens don't exist — there's no dashboard, campaign builder, sequence-generation page, or email editor. What imported cleanly: the AI engine (sequenceGenerator.js calling built-in InvokeLLM), 8 rich entities, the settings pages, and auth. So Sendji isn't a "wire it up" fix like WallCraft/Momentum — its whole main UI has to be built on top of the working engine. That makes it the heaviest of the batch.

ACTIVE MODES — RE-INVOKE THESE FIRST (MANDATORY)

• Was running in fable-mode (five-gate discipline: scope → evidence → adversarial → verify → report). Every prompt in this session was grounded by opening the real Base44 app (entities + pages + grep) before writing — re-invoke fable-mode first if continuing.
• ponytail (full) active via hook — terse, lazy-correct, shortest working output.
• Note: the Fable BUILD prompts themselves instruct each fresh session to invoke base44-workflow + lovable-to-base44-migration (+ fable-mode).

Where it started

Sondra: "we have a list of apps stuck inside Base44, convert them one by one to working apps — grab the list you can see." She interrupted to clarify the stuck ones = the "Migrated from Lovable" shells. Goal became: decode those identically-named shells, then produce paste-ready Fable 5 prompts she fires in parallel sessions to convert each. Mid-session she added a separate ask: fix the original LeadScout Base44 app (specific bugs she hit today).

Decisions locked + what shipped

• Decoded all 7 "Migrated from Lovable" shells (all carry the identical generic name; identified by reading each app's index.html/package.json/entities):
• Momentum 6a4a7a14e5bc05b1006e01d9 · WallCraft 6a4a7a35339244ffc010a9bf · WallCraft DUPE 6a4ac9c6bee3028cccd84370 · Sendji 6a4a7b112b17227e4c8cb654 (title lost in import) · Big Worm 6a4a7b8cf689b9007e734bf5 · Appreneur Challenge 6a52885f4146ba77226eac8a (extra, newer batch) · SpyBench 6a5287dbf6d0b621abf91704 (extra, newer batch).
• She expected 5 (4 templates + 1 dupe). Reality = 7: her 5 (older ~Jul 5 batch) + 2 extras from a newer ~Jul 10-11 batch (Appreneur Challenge, SpyBench).
• Per-app grounded diagnoses (the "don't re-derive" intel):
• WallCraft — fully client-side on Base44 built-in integrations (Core.GenerateImage/UploadFile), NO missing functions. Stuck-points: a Lovable-leftover custom User entity (all fields required) colliding with built-in User + separate Profile; RLS on owner entities; never published; no seed data; generic name.
• Momentum — grep for any LLM/integration call = empty; its headline Coach.jsx (16KB) is likely a stubbed/dead AI feature (wire Core.InvokeLLM). User entity already clean. Plus RLS + rename + seed + publish.
• Sendji — App.jsx routing EMPTY (every URL → PageNotFound); core UI screens (dashboard/campaign builder/sequence gen/email editor) missing, must be built on the working engine. Engine + 8 entities + settings pages + auth imported clean. White-label via AppConfig (indigo #4F46E5 / amber #D97706). Heaviest of the batch.
• LeadScout (original) 6a46ad7a89bffa371e7637da — 3 bugs mapped to code: (1) no campaign-detail/results page exists; (2) Home dashboard cards not clickable; (3) Campaign requires zip_codes → make location accept city/county/zip (Outscraper takes free-text). Live app with REAL data (2 campaigns / 133 leads) — preserve it.
• Shipped: 5 files (all Fable 5, effort high; each tells the fresh session to invoke base44-workflow + lovable-to-base44-migration, sandbox-edit so clones inherit, prove 2-account isolation vs the RLS shorthand trap, keep the app's own brand, not real-money):
• c:\Users\srver\AIOS-Workspace\aios-starter-kit\PROJECTS\Base44 Lovable Conversions\README.md (decoded tracker of all 7)
• ...\PROJECTS\Base44 Lovable Conversions\wallcraft-fable-prompt.md
• ...\PROJECTS\Base44 Lovable Conversions\momentum-fable-prompt.md
• ...\PROJECTS\Base44 Lovable Conversions\sendji-fable-prompt.md
• c:\Users\srver\AIOS-Workspace\aios-starter-kit\PROJECTS\LeadScout Studio\base44-leadscout-original-fix-fable-prompt.md
• Sondra confirmed she'd fire WallCraft + Momentum in parallel sessions; Sendji + LeadScout prompts delivered on request.

Key files for next session

• c:\Users\srver\AIOS-Workspace\aios-starter-kit\PROJECTS\Base44 Lovable Conversions\README.md — the tracker: all 7 shells, IDs, real names, status, open decisions. Read first.
• The 4 prompt files above — paste-ready, one per fresh Fable session.
• No plan file. Memory files touched: none (per skill rules).

Running state

• Background processes: none
• Dev servers / ports: none
• Open worktrees / branches: none

Verification — how to confirm things still work

• python scripts/get_env.py list not needed. To re-list the shells: MCP list_user_apps (Base44), or read the README tracker.
• Each Fable prompt's own "Done =" is the acceptance test (run the app end-to-end, prove 2-account isolation, publish + Remixability ON).

Deferred + open questions

• Deferred: Big Worm 6a4a7b8cf689b9007e734bf5 prompt not yet written (last of the 4 templates) — ground it the same way (pages/entities/grep) then write.
• Open: The 2 extras (Appreneur Challenge 6a52885f..., SpyBench 6a5287db...) — convert them too, or delete? Sondra's call.
• Open: confirm retiring the WallCraft dupe 6a4ac9c6... (plan: keep 6a4a7a35...).

Pick up here

Re-invoke fable-mode first. Then, on Sondra's word: write the Big Worm Fable prompt (same grounded pattern), and/or resolve the 2 open decisions (extras convert-or-delete, WallCraft dupe retire). The 4 delivered prompts run in their own parallel Fable sessions — nothing to do here for those.