Loading ToolBench
Knowledge Shelf MCP Server — on-demand documentation, templates, and workflow automation for AI
Knowledge Shelf provides 7 tools with basic schemas and descriptions, but falls short of production quality. All tools have descriptions (10-80 chars, mostly under the 194-char baseline), and input schemas are present with type definitions. However, descriptions are terse and lack context for LLM selection. Parameter descriptions are minimal or absent. Output schemas are not documented. Error handling is not visible in the code. The run_workflow tool (WRITE operation) lacks confirmation/dry-run patterns. No tool annotations (readOnlyHint/destructiveHint) despite clear risk distinctions. Tool names follow verb_noun convention (good), but parameter naming lacks consistency (e.g., 'name' is overloaded across tools for different resource types). No pagination support visible for list_docs or search_docs despite potentially large result sets.
Retrieve doc content (manifest "doc" field or legacy .md)
Retrieve a specific section from a document
Retrieve manifest.json structured metadata
Retrieve any file from a knowledge folder
List all available knowledge documents/units
Execute a predefined workflow (copy, run, mkdir steps)
Search by keyword (title, tags, description, content)
Descriptions are terse (10-80 chars) and lack context for LLM tool selection. Baseline is 194 chars. E.g., 'List all available knowledge documents/units' does not explain WHEN to call this vs search_docs, or what structure is returned.
Parameter descriptions are missing or minimal. 'name' parameter appears in 5 tools but is never described (e.g., is it a file path, folder name, or document ID?). 'values' in run_workflow lacks any guidance on expected structure or format.
Output schemas are not documented. LLMs cannot plan downstream calls or extract required fields. E.g., does list_docs return an array of objects? What fields does each object contain? Are there pagination cursors?
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
run_workflow (WRITE operation) lacks confirmation/dry-run pattern. Agents can execute arbitrary workflows without preview or rollback. No error recovery guidance if a workflow step fails.
No tool annotations (readOnlyHint/destructiveHint/idempotentHint) despite clear risk distinctions. 6 tools are READ_ONLY; 1 is WRITE. Annotations enable agents to reason about safety and retry logic.