Tools have JSON Schema inputSchema with types, required fields, and descriptions

Without a JSON Schema inputSchema, AI clients have no contract for what arguments a tool accepts — they guess at parameter names, types, and format. OWASP LLM09 (Misinformation) is the direct risk: the AI constructs plausible-looking but wrong invocations, producing silently incorrect behavior. Missing required-field markers mean the AI may omit critical parameters; missing descriptions mean it cannot distinguish a file path from a query string. In prompt-injection scenarios (CWE-20, OWASP A03), undefined inputs widen the attack surface because the server cannot reject malformed or adversarial arguments before execution.

Add a Zod schema as the third argument to every server.tool() call in your TypeScript MCP server. The SDK converts it to inputSchema automatically.

// src/tools/search.ts
server.tool(
  'search_files',
  'Search for files matching a glob pattern',
  {
    pattern: z.string().describe('Glob pattern, e.g. "**/*.ts"'),
    directory: z.string().optional().describe('Root directory to search; defaults to project root'),
    maxResults: z.number().int().min(1).max(500).default(50),
  },
  async ({ pattern, directory, maxResults }) => {
    // handler
  }
)

For Python, add type annotations and a docstring with Args: block — the SDK derives the schema from them automatically.