examples
Client Role Grants (anon/authenticated) examples
These examples show how Client Role Grants (anon/authenticated) problems ship in real apps — and what fixes actually work when tested via direct API access.
Why Client Role Grants (anon/authenticated) examples matter
You lose control over who sees or changes data, which is the root cause of many Supabase leaks. If Client Role Grants (anon/authenticated) remains unresolved, attackers can automate enumeration and unauthorized writes at API speed. Treat it as a production reliability risk as well as a data security risk, because incidents spread quickly once clients discover weak access boundaries.
Examples about Client Role Grants (anon/authenticated)
| Example | Summary | URL |
|---|---|---|
| Authenticated SELECT on a sensitive table | Authenticated SELECT turned a sensitive table into an open endpoint. | /examples/client-role-grants/authenticated-select-on-sensitive-table |
| View grant exposes sensitive columns | Granting a convenience view to client roles exposed sensitive columns. | /examples/client-role-grants/view-grant-exposes-sensitive-columns |
Root cause → fix pattern analysis for Client Role Grants (anon/authenticated)
Examples are most useful when you can translate them into a repeatable fix pattern. This table highlights the “why” behind each fix:
| Example | Root cause | Fix pattern | URL |
|---|---|---|---|
| Authenticated SELECT on a sensitive table | The grant made the table readable by every logged-in user, allowing enumeration when the policy was missing or permissive. | Revoke SELECT from client roles, expose the feed through a backend endpoint, and verify direct REST calls fail. | /examples/client-role-grants/authenticated-select-on-sensitive-table |
| View grant exposes sensitive columns | The view began returning billing data and emails; attackers queried it directly and retrieved columns that were never meant for the browser. | Revoke client access, replace it with a backend endpoint that returns a minimal, documented shape, and treat views/functions as API surfaces in your audits. | /examples/client-role-grants/view-grant-exposes-sensitive-columns |
How Client Role Grants (anon/authenticated) failures typically happen
- Granting SELECT to client roles for convenience.
- Forgetting to revoke INSERT/UPDATE after policies land.
- Trusting UI gating instead of database restrictions.
Fix patterns that tend to work for Client Role Grants (anon/authenticated)
Across these examples, the highest-leverage fixes share a theme: remove direct client access and make verification repeatable.
- Backend-only access for sensitive operations (server endpoints enforce authorization).
- Least-privilege grants: revoke broad privileges from anon/authenticated.
- Small, testable policies if you intentionally keep client access — avoid complex conditions.
- A verification step that proves direct access fails (not just that the UI hides data).
How to spot Client Role Grants (anon/authenticated) in your own project (signals)
- A direct API call returns rows/files even when the UI is supposed to restrict them.
- RLS/policies exist, but access still succeeds (often because RLS is disabled or policies are too broad).
- Permissions depend on the client behaving “nicely” (UI checks) rather than the database enforcing access.
- After a migration, access behavior changes unexpectedly (drift).
How to use these examples to fix your own app
- Match the scenario to your table/bucket/function setup.
- Identify the root cause (not just the symptom).
- Apply the relevant template or conversion guide.
- Verify direct access fails for client credentials.
- Document the rule so it doesn’t regress.
Verification checklist for Client Role Grants (anon/authenticated) fixes
- Reproduce the issue once using direct API access (anon/authenticated) so you know it’s real.
- Apply the fix pattern (backend-only access + least privilege) using a template.
- Repeat the same direct access call and confirm it now fails.
- Confirm the app still works via backend endpoints for authorized users.
- Re-scan after the fix and add a drift guard for the next migration.
Preventing Client Role Grants (anon/authenticated) regressions (drift guard)
- Re-run the same direct access test after every migration that touches auth, policies, grants, Storage, or functions.
- Keep a short inventory of sensitive resources and treat them as server-only by default.
- Review new tables/buckets/functions in code review with an access-control checklist.
- If you intentionally allow client access, document the policy rationale and add tests for it.
Optional SQL checks for Client Role Grants (anon/authenticated) (extra confidence)
If you like having a repeatable “proof”, add a small set of SQL checks to your process.
- Confirm RLS status for tables involved (enabled/forced where appropriate).
- List policies and read them as plain language: who can do what under what condition?
- Audit grants to anon/authenticated and PUBLIC for tables, views, and functions tied to this topic.
- If Storage/RPC is involved, explicitly audit bucket settings and EXECUTE grants.
These checks complement (not replace) the direct access tests shown in the examples.
Decision guide for Client Role Grants (anon/authenticated): template vs conversion vs integration
If you’re here because you found this topic in a scan, the fastest path depends on whether the fix is a small config change or a workflow change.
- Choose a template when you need a copy/paste change plus verification (tighten a policy/grant/bucket setting).
- Choose a conversion when you need to change an access model end-to-end (unsafe → backend-only) with example transformations.
- Choose an integration when the fix is a workflow pattern you’ll repeat (signed URLs, server-only RPC, backend endpoints).
If you’re unsure, start with the smallest template that removes direct client access, then add integrations for durability.
Evidence to keep after fixing Client Role Grants (anon/authenticated) (makes reviews faster)
Teams often “fix” a topic but can’t prove it later. Save a few small artifacts so you can re-verify after migrations:
- The direct access request you used before the fix (and the expected denial after).
- A short boundary statement (who can access what, through which server endpoint).
- The change you applied (policy/grant/bucket setting/EXECUTE revoke) and why.
- The drift guard you’ll run after migrations (scan, checklist query, or release checklist item).
Related pages
- Glossary: Client Role Grants (anon/authenticated) →
/glossary/client-role-grants - Template: Lock down a public table (backend-only access) →
/templates/access-control/lock-down-public-table - Template: Lock down RPC: revoke EXECUTE from public roles →
/templates/rpc-functions/lock-down-rpc-execute
What to do after you fix one example (so it stays fixed)
One fixed example is great — but the real win is preventing drift.
- Write a one-sentence boundary statement (who can access what, through which server path).
- Keep the one direct access test you used before the fix (and expect it to fail after).
- Re-run the same test after migrations that touch policies, grants, buckets, or functions.
If you can re-run the test and it still fails, you’ve turned a one-time fix into a durable control.
FAQ
What’s the fastest fix pattern when Client Role Grants (anon/authenticated) shows up in a scan?
Prefer backend-only access and remove direct client privileges. Then add verification checks that prove direct access fails.
Can I fix Client Role Grants (anon/authenticated) with policies alone?
Sometimes, but it’s easy to get subtly wrong. Use these examples to learn the failure modes, and verify with direct API tests.
How do I choose between examples, templates, and conversions?
Examples explain the pattern, templates show concrete implementation, and conversions describe the whole transformation from unsafe to safe.
Next step
Want to know if your project matches any of these scenarios? Run a Mockly scan and compare your findings to the examples here.