musterd universal harness adapter: one MCP server that joins any MCP-capable agent to a team. Provides coordination tools for multi-agent teams with presence, inbox, memory, goals, and work-board (lane) management.
Critical failures across all dimensions. Tool definitions exist only as string constants in toolNames.ts with no visible schema registration, parameter definitions, or output documentation. No input schemas found in source. Descriptions are minimal (10-30 chars), far below the 50-200 char baseline for LLM-optimized tools. No error handling guidance, no parameter constraints, no composition support. Tools appear to be name-only stubs without implementation details visible in provided source.
Get the work board status
Claim a work lane
Hand off a work lane to another team member
Open a new work lane
Release a claimed work lane
Resolve a work lane
Submit a work lane for review
Update a work lane
No input schemas visible for any tool. All 29 tools lack parameter definitions, type constraints, and validation rules. Cannot verify LLM can invoke tools correctly.
Descriptions are minimal (10-30 chars), far below 50-200 char baseline. Examples: 'Join a team as a member' (26 chars), 'Leave a team' (12 chars). LLMs cannot distinguish when to use similar tools or understand prerequisites.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 40 | 2026-07-28+ | v2 |
Check team member availability
Declare a team goal
Report the outcome of a team goal
Retract a team goal
List team goals
Check team inbox for messages
Save an insight to team memory
Search team insights
Join a team as a member
Leave a team
List team members
Read information from team memory
Save information to team memory
Get the next action for the team
Generate a team report
Get a specific team seed
List team seeds
Update a team seed
Send a message to the team
Get team status
Get context for waking a team member
No output schemas documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., team_id, lane_id) for composition.
No error handling guidance. Tools lack recovery hints, retryability classification, or actionable error messages. LLMs cannot self-correct on failures.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). LLMs cannot determine which tools are safe to retry or have irreversible side effects.