Korean NLP Tools server has 5 tools with basic schemas and Korean-language descriptions. All tools have input schemas with type definitions and parameter descriptions. However, descriptions are minimal (10-50 chars), lack context on WHEN to use each tool, and do not explain output structure. No output schemas documented. Tool names lack English verb prefixes (tokenize is acceptable, but checkSpelling, normalizeText, extractKeywords, fixSpacing are not verb_noun format). No error handling guidance, no pagination support, no composition hints. All tools are READ_ONLY, which is good, but descriptions do not state this explicitly. Baseline: avg tool description ~25 chars (well below 194-char production baseline); zero output schema documentation.

• Expand tool descriptions to 100-200 chars. Example: 'tokenize: Morphologically analyze Korean text, splitting into parts-of-speech (noun, verb, particle, ending). Returns array of {text, pos, lemma}. Use before keyword extraction or sentiment analysis.'
• Document output schemas for all tools. Show example responses: tokenize returns [{text: string, pos: string}], checkSpelling returns {errors: [{position, suggestion}], is_correct: boolean}.
• Rename tools to verb_noun format: check_spelling, normalize_text, extract_keywords, fix_spacing. This matches production baselines and helps LLMs parse intent from names alone.
• Add error handling guidance: 'If text is empty, returns empty array. If text contains non-Korean characters, they are passed through unchanged. Invalid UTF-8 returns error: "Invalid encoding".'
• For extractKeywords, document result limits: 'Returns up to top_n keywords (default 10, max 100). If fewer than top_n keywords meet min_length threshold, returns all available.'