The same access rule created over and over
Scan rule · RSC-ARC-021
- Architecture & SEO
- Severity: Medium
Finds RLS policies recreated three or more times across the migrations (each after a `drop policy`) — the sign that the AI is spiralling on an access rule instead of setting it correctly once.
Why this matters
In vibe coding the AI proposes a policy, it does not fit, the next migration drops it and creates a new one — repeatedly. Recreating the same policy three times is no longer iteration but spiralling: the rule was never understood as a whole.
In the field corpus, repos with this policy churn carried a critical RLS gap far more often than repos without — the last version is often a quick fix that still leaves the door open.
Fix: step back and ask what the same access logic looks like done cleanly — usually one policy per operation with `auth.uid() = user_id`. That is shorter and safer than the next rewrite.
Scan tier
- Repo access (deep scan)
How to fix it
State the access rule cleanly once: one policy per operation (select/insert/update/delete), owner bound via `auth.uid()` — instead of rewriting the same policy again.