Read-only Postgres roles aren't truly read-only; dev open-sources scanner
Then_Respect_1964 · reddit · 2026-10-04
A developer points out that handing "read-only" Postgres credentials to MCP/agent workloads is unreliable: a role can gain privileges you didn't explicitly grant through inherited roles, ownership, PUBLIC grants, and more. What matters is the credential's effective permissions, not how it was originally provisioned.
To address this, the author open-sourced agent-db-scan: give it a Postgres connection string and it shows what that login can actually do. The author had been using manual scripting internally to check this before handing credentials to agents, and released a small open-source version. The project is still early.
More from coding & agent
- Netlify adds GPT-6.1 Sol to AI Gateway and Agent Runners with zero-config auth — jasonkneen · 2026-10-04
- Hot take: game dev is the hardest software — agents will crack it within 6 months — repligate · 2026-10-04
- Swarms Rust multi-agent framework launches: 6ms cold start, up to 440x faster than LangGraph — KyeGomezB · 2026-10-04
- CMU open-sources Quail, an AI-SQL engine that lets LLMs filter rows in SQL — sh_reya · 2026-10-04
- Dev builds Pix4less with Claude Code — a 30k-image cache layer so AI agents search before generating — No-Line-3463 · 2026-10-04
- call-me MCP upgrade: agent phone calls now read 2-minute morning briefings via AWS Polly — radressss · 2026-10-04