Instruction file imported from Sinedrone-Sentinel/dumpers_repo (
.cursor/rules/missed-setup-steps.mdc). Copyright stays with the author.
Missed Setup Steps - Tell, Don't Bypass
When an error, failure, or odd state is explained by the user (or environment) skipping a required step, say that clearly. Do not rewrite app code, migrations, or RPCs to paper over the miss.
Hard rule
- Diagnose whether the fix is "run/apply/merge X first" (migration order, env var, deploy, PR merge, script, dependency).
- If yes -> tell the user the missed step (what, why, exact order). Stop there unless they ask you to change code.
- Do not invent workarounds that duplicate helpers, inline prior migration logic, or weaken dependencies so an incomplete setup "still works."
Especially (Supabase / this repo)
Migrations are ordered. If SQL fails with function/type/relation does not exist and that object is created in an earlier migration file:
- Say: apply migration N before N+1 (list the files).
- Do not copy helpers from an earlier migration into the later one just so it runs alone.
Same for: Edge Function not deployed, service-role env missing, PR not merged, npm script not run, parse/extract step skipped.
Good vs bad
BAD: Error "friend_rsi_label does not exist" -> rewrite 169 to avoid calling it / redefine it inline
GOOD: "169 needs 168 first - apply 168_friends_rsi_handle_privacy.sql, then re-run 169."
BAD: Feature half-broken because RPC not deployed -> change client to hit tables directly
GOOD: "Deploy / apply migration X; client already calls the right RPC."When coding is appropriate
Only change code when the repo is wrong (bug, real missing dependency in git, bad migration that can't be applied in order even when prior steps ran). If prior steps weren't applied, that is not a code bug - report the gap.