AI article

Your MCP server's tool descriptions are an attack surface

Community description: MCP tool descriptions are text. When your agent calls tools/list, the server returns JSON with a...

Dev.to | Mar 5, 2026 | LPW

Automated excerpt

When your agent calls tools/list, the server returns JSON with a description field for each tool. Their "Full Schema Poisoning" research tested multiple fields: Nested description injection. Instructions hidden in description fields inside the inputSchema properties, not in the top-level tool description: The top-level description is clean.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.

Read the original article

More AI news