Vale offers 4 well-named tools with complete input schemas and clear descriptions. Tool names follow verb_noun convention (lint_text, list_rules, fix_text, update_vocabulary). All tools have descriptions (avg 140 chars, within 10-1024 baseline). Input schemas are properly typed with enums where appropriate (minSeverity). However, output schemas are completely undocumented, LLMs cannot plan downstream calls or extract structured data. Parameter descriptions are present but minimal (avg 45 chars). Error handling and recovery guidance are absent from visible code. The fix_text tool accepts model/temperature parameters but lacks documentation on defaults, constraints, or failure modes. update_vocabulary persists state but lacks documentation on scope (session vs persistent) or rollback behavior.

• Document output schemas for all tools. E.g., lint_text should return {findings: [{rule, severity, message, line, column}], summary: {errorCount, warningCount}}. This enables LLMs to extract and chain results.
• Add constraints to fix_text parameters: document valid model names (e.g., 'gpt-4, gpt-3.5-turbo'), temperature range (0.0-2.0), and maxTokens bounds (1-4096). Include defaults in description.
• Clarify update_vocabulary scope: does it persist across sessions? Can concurrent calls race? Add example: 'allow: ["API", "REST"] to stop flagging these terms in subsequent lint_text calls.'
• Add error recovery guidance to descriptions. E.g., fix_text: 'If the model API is unavailable, lint_text will still work. Retry fix_text after checking service status.'
• Document minSeverity filtering behavior in lint_text: does 'warning' return warnings + errors, or warnings only? Clarify enum semantics.