Demo MCP servers for the aMaze Control Plane platform, including tools for SQL queries, web search, and PDF processing
Single tool 'sql_query' has a reasonable description (194 chars) explaining the database schema and supported operations. However, the input schema is critically incomplete: only a 'query' parameter with type 'string' and basic description is visible. No constraints (regex, length limits, enum values for query types), no output schema documentation, and no error handling guidance. The tool accepts WRITE operations (INSERT/UPDATE/DELETE) without confirmation, dry-run, or permission checks, a critical security gap. Description lacks WHEN to use this tool vs alternatives, prerequisites, or recovery guidance for common errors. No parameter validation rules stated (e.g., max query length, timeout behavior). Tool does not follow the pattern:command-tool or pattern:confirmation-request for destructive operations.
Execute a SQL query against the MySQL database and return the results. The database contains three tables: - users(id, name, email, city, age, is_active, phone, credit_card, created_at) - products(id, name, description, price, category, stock, created_at) - orders(id, user_id, product_id, quantity, total_price, status, created_at) Use SELECT queries to explore data. INSERT/UPDATE/DELETE are also supported.
No input schema constraints. 'query' parameter accepts any string with no length limit, regex pattern, or validation rules. LLMs can pass arbitrarily long or malformed SQL, causing timeouts or injection attacks.
No output schema documented. Tool description mentions three tables but does not specify what fields are returned, data types, pagination behavior, or result limits. LLMs cannot plan downstream operations without knowing the response structure.
Destructive operations (INSERT/UPDATE/DELETE) lack confirmation or dry-run support. Tool accepts WRITE queries without permission checks or user confirmation, violating pattern:confirmation-request. A single malformed agent call could corrupt the database.
No error handling guidance. Description does not explain what happens on SQL syntax errors, connection failures, permission denials, or timeouts. LLMs receive raw database errors with no recovery path.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | <=2025-11-25 | v2 |
Parameter description lacks actionable constraints. 'SQL query to execute' does not specify format, length limits, allowed statement types (SELECT only vs all), or examples of valid queries. Ambiguous parameter descriptions force LLMs to guess.