Server defines 14 tools with complete JSON schemas and descriptions in Japanese. Naming follows verb_noun pattern (get_, search_) consistently. However, critical gaps exist: (1) Descriptions are in Japanese only, reducing accessibility for LLM context in English-dominant environments; (2) No output schemas are documented, LLMs cannot determine what fields to expect from tool responses; (3) Error handling is minimal, no recovery guidance, no actionable error messages; (4) Parameter descriptions lack constraint details (e.g., 'limit' has min/max but no explanation of what happens if exceeded); (5) No pagination guidance despite several tools accepting 'limit' parameters; (6) Tool compositions assume multi-call patterns (e.g., get_channel_info then get_stream_info) without chaining IDs documented in responses. Average tool score: 54/100.
チャンネル情報を取得します
チャット設定を取得します
クリップの情報を取得します
特定のゲームの情報を取得します
グローバルチャットバッジのリストを取得します
グローバルエモートのリストを取得します
配信情報を取得します
No output schemas documented. Tool descriptions specify inputs but not return types. LLMs cannot plan downstream tool calls or extract chaining IDs (e.g., channel_id for follow-up queries). This forces agents to treat responses as unstructured text.
All tool descriptions are in Japanese only (e.g., 'チャンネル情報を取得します', 'ユーザーの情報を取得します'). English-language LLMs and tool discovery systems cannot understand purpose or selection criteria. Descriptions should be in English or include English translations.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 45 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 43 | - | v1 |
現在ライブ配信中のストリームを取得します
人気のゲームのリストを取得します
ユーザーの情報を取得します
アーカイブ動画のコメントを取得します
チャンネルのビデオを取得します
ゲームやカテゴリーを検索します
チャンネルを検索します
Parameter descriptions lack actionable constraint details. Example: 'limit' parameter in get_top_games, search_categories, search_channels, get_streams, get_clips, get_videos, get_video_comments has min/max in schema (1 - 100) but descriptions do not explain behavior at boundaries or what happens if limit exceeds available results. LLMs cannot self-validate inputs without explicit descriptions of constraints.
No pagination guidance or next_cursor/page tokens documented. Tools like get_top_games, search_categories, search_channels, get_streams accept limit but no offset or cursor parameter. If an agent needs results 21 - 40, it cannot page, it must either accept partial results or retry with a different query. This blocks multi-page discovery patterns.
Error handling is minimal. The server catches errors in src/index.ts (handleApiError) but source does not show actionable error messages. LLMs receive raw API errors without recovery guidance (e.g., 'User not found. Try search_users() with a partial name'). Agents cannot self-correct on 404, auth, or rate-limit errors.
Tool composition assumes agents make multiple sequential calls without chaining IDs documented. Example: get_channel_info returns channel data, but no documented output schema confirms it includes a channel_id for use in get_stream_info, get_clips, get_videos. Breaking chaining forces discovery detours and wastes context.
Parameter names lack type suffixes. Example: 'channelName' in get_channel_info, get_stream_info, get_chat_settings, get_clips, get_videos, is this a string display name or an internal channel ID? Clarity would benefit from 'channel_name' vs 'channel_id' distinction or a description stating 'Twitch channel name (e.g., "twitchdev").' Current naming ambiguity forces LLMs to guess or make extra lookups.
Tools returning lists (get_top_games, search_categories, search_channels, get_streams, get_global_emotes, get_global_badges, get_clips, get_videos) do not document a total count or next_cursor. Without this metadata, agents cannot know if more results exist or how many total items matched the query. Forced to either accept partial results or guess pagination strategy.