DESIGN MEMORY: recorded decisions/logs/rejections with who/when. Read BEFORE changing a system's direction; empty = no recorded WHY — don't invent one.
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | Default 30, max 100 | |
| query | No | Keyword filter | |
| system | No | System id or exact name | |
| project_id | No |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The readOnlyHint annotation already marks the operation as safe, but the description adds valuable behavioral context: it describes the content (decisions/logs/rejections), the associated metadata (who/when), and the meaning of an empty result. The warning 'don't invent one' is a critical behavioral caveat that goes beyond the annotation. No contradiction.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences, front-loaded with a 'DESIGN MEMORY' label, and every word adds value. It covers content, usage, and empty-result semantics without fluff. This is exemplary conciseness.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has 4 optional parameters, no output schema, and a single annotation, the description provides strong contextual coverage: it explains what the tool returns (who/when), when to use it, and how to interpret empty results. It does not detail return format or parameter relationships, but for a simple read tool, it is sufficiently complete. A slightly higher score would require more specific output structure or parameter interaction details.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema already describes 3 of 4 parameters (limit, query, system) with useful details like defaults and formats. The description does not mention parameters, but with 75% schema coverage, it doesn't need to compensate. The 'system' parameter is indirectly tied to the usage text ('changing a system's direction'), but not explicitly explained. Overall, the description adds little parameter meaning, but the schema mostly covers it.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly identifies the tool as retrieving design memory—recorded decisions, logs, and rejections with who/when. The phrase 'DESIGN MEMORY' and the explicit mention of 'recorded decisions/logs/rejections' distinguishes it from sibling tools like get_rejections, which focuses only on rejections. The purpose is specific and actionable.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description states a clear usage trigger: 'Read BEFORE changing a system's direction.' This tells the agent when to invoke the tool. It also adds interpretive guidance ('empty = no recorded WHY — don't invent one'), but it does not name alternatives or explicitly state when not to use it. Still, the guidance is strong and context-rich.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Add one secure layer between your agents and this server.
Each tool has a clearly distinct purpose, with explicit distinctions between direct actions and proposals via Inbox. The verbs and object types (system, milestone, screen, element, balance) are unique enough that no two tools appear to do the same thing.
Most tool names follow a consistent verb_noun snake_case pattern (get_system, propose_screen, update_element). Minor deviations like 'dedupe', 'search', 'next_task', and 'reorder' are single words or non-verb but remain readable and stylistically compatible.
With 54 tools, this server vastly exceeds the typical MCP scope, hitting the 'extreme mismatch' threshold. Even for a complex domain, the sheer number will overwhelm agents and degrade selection performance.
The tool surface is remarkably complete, covering full lifecycle operations for all major entities, plus import, design generation, drift detection, status reporting, inbox handling, and rejection workflows. No obvious dead ends or missing operations for the stated purpose.