MCP server for workflow execution, step management, and knowledge base operations with SQLite backend
Four tools with complete input schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools have descriptions (10-100+ chars), but 3 of 4 lack descriptions for nested object properties. workflow.next_step has a complex nested schema with undocumented artifact and finding structures. No output schemas documented. Error handling is absent, tools do not guide recovery or categorize failures. Parameter descriptions exist but lack format constraints, ranges, or validation rules. Naming follows verb_noun pattern (workflow.*, knowledge.*) but lacks clarity on side effects and prerequisites.
Store a long-lived finding with optional project scope.
Update an existing knowledge finding.
Continue workflow to next step. Submit step output and receive next agent persona.
Start a new workflow execution.
workflow.next_step has deeply nested artifact and finding objects with no descriptions for inner properties (type, title, name, content, content_type, metadata, scope, category, severity, tags, source_execution_id, source_agent). LLMs cannot infer what these fields control or how to populate them.
No output schemas documented for any tool. Callers do not know what fields to expect in responses, forcing LLMs to guess at downstream field names and risk broken tool chains.
No error handling guidance. Tools do not indicate which errors are retryable, what caused them, or what the LLM should do next. A failed workflow.next_step call leaves the agent stuck.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2025-06-18+ | v2 |
Parameter descriptions lack format constraints and validation rules. E.g., 'token' in workflow.next_step has no format hint; 'project_id' in knowledge.add_finding has no range (is it 1-2^31?); 'severity' enum is documented but not explained (what distinguishes 'low' from 'medium'?).
workflow.next_step description does not clarify that this is a stateful operation with side effects (persists findings, advances workflow state). Agents need to know this is not idempotent and cannot be safely retried.