@benborla29/mcp-server-mysql

SQL injection vulnerability in executeQuery

src/db/index.ts:54 - executeQuery accepts raw SQL with params array but doesn't validate input. Multiple callers pass unsanitized user input directly to connection.query(sql, params). No input validation on sql parameter before execution.

Hardcoded credentials in repository

.env file committed with credentials: MYSQL_PASS='my_secure_password', MYSQL_USER='root'. .env.test contains actual test credentials. Both files are tracked in repo (1309 and 1285 bytes respectively).

Command injection via MYSQL_CONNECTION_STRING

src/utils/index.ts:68-157 parseMySQLConnectionString() parses raw connection strings without validation. Accepts arbitrary mysql command-line arguments. No sanitization of socketPath, host, or other parameters before use in mysql2.createPool().

Insufficient path traversal protection

src/utils/index.ts:129 - socketPath can be set via connection string or env var with no validation. Could reference arbitrary filesystem paths: config.socketPath = value (no path sanitization, no chroot checks).

Multi-statement execution enabled by default

src/config/index.ts and scripts/setup-test-db.ts - multipleStatements: true in connection config. Allows SQL injection to execute multiple queries in single call, escalating impact of any injection vulnerability.

Permission bypass in multi-DB mode

src/db/permissions.ts and src/db/utils.ts - extractSchemaFromQuery() uses regex to parse schema names. Can be bypassed with crafted queries using comments or alternate syntax. isInsertAllowedForSchema() returns global default if schema extraction fails.

Verbose error messages leak system information

No rate limiting or query complexity checks

Weak password handling