Loading ToolBench
AI Learning & Accountability System for SPARK PKM - FastAPI backend with MCP client integration for vault access, learning resource management, briefing generation, quizzes, nudges, voice processing, and statistics
This MCP server has severe definition quality gaps. All 4 tools have minimal or missing descriptions, no documented input schemas beyond what can be inferred, and poor parameter documentation. The tools are not LLM-optimized and lack the clarity needed for reliable agent selection. Only 2 of 4 tools have any meaningful input parameters at all. The server provides no pagination, error recovery guidance, or output schema documentation. Naming conventions are weak ('test_mcp_*' violates the verb_noun_resource pattern and uses 'test' which is vague). This is well below production baseline.
Health check endpoint - no authentication required. Returns 200 if the API is running
Test MCP server connectivity by reading a note from the vault. Requires authentication. Returns sample note from the vault or connection status
Test reading a specific note by path. Requires authentication. Returns note content and metadata
Test MCP search functionality. Requires authentication. Returns search results from vault
Tool names violate verb_noun pattern. All three 'test_mcp_*' tools use 'test' prefix which is vague and generic, making it unclear what action is performed. Should be 'read_note', 'search_vault', 'get_note' etc.
Parameter descriptions are minimal or missing context. 'query' in test_mcp_search lacks detail about required format, query syntax, supported operators, or example syntax. Same for 'path' and 'folder' in test_mcp_read and test_mcp_search. Parameters need to describe expected format, range, and dependencies per pattern:tool-description.
No output schema documentation. Tools return responses but the structure, field types, pagination, and chaining fields are not documented. LLMs cannot plan downstream calls or extract the correct data without documented return schemas.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 48 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 48 | - | v1 |
No pagination support. test_mcp_search returns 'search results from vault' with no indication of limit, offset, page size, or total count. Large result sets can exhaust context window without pagination guidance.
Descriptions are below 50 character baseline for clarity. 'Health check endpoint - no authentication required. Returns 200 if the API is running' (72 chars) is acceptable, but 'Test MCP server connectivity by reading a note from the vault. Requires authentication. Returns sample note from the vault or connection status' (138 chars) is verbose and doesn't clearly state WHEN to use vs other tools. Descriptions should state what distinguishes each tool from alternatives.
No error handling guidance. Descriptions do not explain what failures mean or how agents should recover. 'Requires authentication' indicates auth is needed but doesn't tell the agent how to handle auth failures.
Generic tool names confuse similar operations. 'test_mcp_search', 'test_mcp_read', and 'test_mcp_connection' are all read-only vault operations but LLMs cannot easily distinguish their use cases. Consider: search_vault_notes, read_vault_note, verify_vault_access.
Parameter 'folder' in test_mcp_search is optional but no default or behavior is documented. Does omitting it search all folders? The root? Specific folders? LLM behavior is undefined.