Loading ToolBench
Model Context Protocol server for Hibiki content creation platform. Enables AI-powered multi-platform social media content creation and management.
Server has 6 tools with complete input schemas and descriptions. Naming follows verb_noun convention (create-post, get-post, update-post). Descriptions are present and adequate (100-200 chars typical). However, output schemas are not documented in the visible code, only input schemas are defined. Parameter descriptions are present but some lack detail on constraints (e.g., 'page' and 'limit' in get-all-posts lack min/max bounds). Error handling is minimal, no recovery guidance or actionable error messages visible. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear read/write semantics. Composition is reasonable: tools are single-purpose and chainable (create-post returns post data that get-post can retrieve). Missing: output schema documentation, error recovery patterns, and tool annotations.
Create a new content post with platform-specific variations. Enhances the content for each platform, optionally generates images, and can schedule or publish to social networks.
Retrieve all posts with pagination and optional filtering by status. Returns a list of posts with their metadata.
Retrieve a specific brand style by ID with full details including colors, tone, and imagery style.
Retrieve all brand styles for the authenticated user. Returns a list of brand styles with their basic information.
Retrieve a post by its ID. Returns the original content, platform-specific variations, generated images, and metadata.
Update an existing post. Can modify title, content, platform-specific variations, and status.
Output schemas not documented. LLMs cannot plan downstream tool calls or extract required fields without knowing response structure. create-post, get-post, update-post, get-all-posts, get-brand-styles, get-brand-style all lack documented return types.
Numeric parameters lack bounds. get-all-posts 'page' and 'limit' have no min/max constraints documented. LLMs may pass absurd values (page=999999, limit=1000000) that break pagination or timeout.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). create-post and update-post are clearly destructive/write operations; get-* tools are read-only. Annotations would help agents reason about safety and retry logic.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
No error recovery guidance. Tool descriptions do not explain what to do if a call fails (e.g., 'If post not found, try get-all-posts to list available posts'). Error responses likely lack actionable next steps.
Parameter descriptions lack format/constraint details. 'scheduleTime' in create-post says 'ISO 8601 datetime' but does not specify timezone handling, past-date rejection, or max future offset. 'brandStyleId' is vague, should it be a UUID, slug, or numeric ID?