AI agents invoke safe_alter_table to trigger actions in Postgres. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call — builds kicked off, notifications sent, workflows started.
ALTER TABLE statements modify database schema (adding/dropping columns, changing constraints, renaming, etc.). While labeled 'safe', this tool executes DDL that can have irreversible effects (e.g., dropping columns) or at minimum schema-modifying effects. Without a full description, the exact scope is unclear, but DDL execution on a live database warrants Execute/high severity.
From the tool's definition 'safe_alter_table' - converts a high-level intent into DDL ALTER TABLE operations on a PostgreSQL database; description is truncated/uninformative beyond the initial phrase
Attacks that exploit this kind of access
Convert a high-level intent (. It is categorised as a Execute tool in the Postgres MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Postgres MCP server in PolicyLayer and add a rule for safe_alter_table: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Postgres. Nothing to install.
safe_alter_table is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the safe_alter_table rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for safe_alter_table. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
safe_alter_table is provided by the Postgres MCP server (teja-sudo/postgres-mcp-server). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →