Server has 41 tools with consistent naming (verb_noun pattern) and complete input schemas. Descriptions are present but brief (avg ~40 chars, well below the 194-char baseline). Parameter descriptions are minimal, most lack detail on constraints, formats, or valid ranges. Output schemas are not documented. Error handling is absent from tool definitions. Security-sensitive tools (delete_auth_user, execute_sql) lack confirmation or dry-run patterns. No tool annotations (readOnlyHint, destructiveHint) despite clear risk stratification in metadata.
Apply a SQL migration and record it. Privileged.
Create a new auth user. Privileged — requires service_key.
Delete an auth user. Privileged.
Execute an arbitrary SQL query. For SELECT/WITH queries, returns result rows. For DDL/DML, returns status. Privileged — requires service_key.
Run EXPLAIN ANALYZE on a query.
Get security and performance advisory information. Checks for common issues: unused indexes, tables without primary keys, tables without RLS in public schema.
Descriptions are uniformly brief (10 - 60 chars) and lack context on WHEN to use each tool, prerequisites, or dependencies. Baseline is 194 chars. LLMs cannot distinguish between similar tools (e.g., list_indexes vs list_constraints) without richer guidance.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | <=2025-11-25 | v2 |
Get details for a specific auth user.
Show recent execution history for a cron job.
Show active database connections from pg_stat_activity.
Retrieve database statistics from pg_stat_database.
Get metadata for an edge function.
Get the source definition of a database function.
Show index usage statistics.
Retrieve general Supabase logs. Note: Log access depends on the self-hosted logging/analytics configuration.
Show RLS enabled/disabled status for all tables in a schema.
Get configuration for a storage bucket.
Get the source definition of a trigger.
Show pgvector index usage statistics.
Invoke an edge function. Privileged.
List auth users.
List all available (installable) PostgreSQL extensions.
List constraints for a specific table.
List scheduled pg_cron jobs (requires pg_cron extension).
List user-defined database functions.
Retrieve recent logs for an edge function. Note: Log access depends on the self-hosted Supabase logging configuration.
List deployed edge functions. Note: Self-hosted Supabase may not expose a function listing API. Falls back to a helpful message if not available.
List installed PostgreSQL extensions.
List foreign keys for a specific table.
List indexes for a specific table.
List applied Supabase migrations.
List PostgreSQL publications used for Supabase Realtime.
List Row-Level Security policies for a table.
List all storage buckets.
List objects in a storage bucket.
List columns for a specific table.
List tables in a database schema.
List triggers for a specific table.
List pgvector indexes (requires pgvector extension).
Restart pg_net worker and notify PostgREST to reload schema. Privileged.
Update an existing auth user. Privileged.
Update storage bucket configuration. Privileged.
Parameter descriptions are minimal or absent. E.g., 'schema' param lacks guidance on default behavior, valid values, or format. 'instance' param appears in all tools but is never explained. LLMs cannot infer constraints from names alone.
No output schemas documented. Tools return data but LLMs cannot plan downstream calls or extract fields without knowing the response structure. E.g., does list_tables return table_name or name? Does it include schema info?
Destructive tools (delete_auth_user, execute_sql, apply_migration) lack confirmation or dry-run patterns. No error guidance on recovery. Agents can accidentally delete users or corrupt schemas without safeguards.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk stratification in metadata. Clients cannot distinguish safe reads from destructive writes without parsing descriptions.