MCP tool poisoning hides instructions inside an AI tool's name, description, or parameter schema, so an agent reading the tool's documentation treats them as legitimate guidance. It matters because agents install tools from barely screened public registries, and most client interfaces never show...
Source: [Dev.to](https://dev.to/ali_82f3409ed625e8ba8515b/mcp-tool-poisoning-build-the-attack-then-block-it-2g9)