A Model Context Protocol server providing access to the live Affine.Earth Math Court, a multi-domain verdict system for fusion control, actuarial analysis, atmospheric monitoring, and other scientific domains. Forwards tool calls to affine.earth membrane.
This server exhibits severe definition quality issues across nearly all dimensions. 14 of 21 tools (67%) have descriptions that are trivial forwarding statements ('Forward X to the live membrane') with zero actionable context. Only 7 tools expose input schemas; 14 tools have no visible schema at all. Tool names lack action verbs and are domain-specific jargon (corpus_bonds, critique_frame, umc_direct) that provide no semantic clarity to LLMs. No output schemas are documented. Error handling is absent. The server appears to be a thin HTTP proxy to an external 'membrane' service, with tool definitions inferred rather than explicitly registered in MCP format. This violates core MCP patterns for tool composition and LLM-friendly interface design.
Forward atc.assert_4d_deconfliction to the live membrane.
Forward corpus_bonds to the live membrane.
POST an integer court claim. Floats are refused by the cell.
Forward critique_frame to the live membrane.
Execute a 2-local Hamiltonian constraint satisfaction problem.
Execute exact permanent computation on a matrix.
Forward execute_transition to the live membrane.
14 of 21 tools (67%) have trivial descriptions ('Forward X to the live membrane') with zero semantic content. LLMs cannot determine when or why to select these tools. Violates pattern:tool-description baseline of 194 chars average and requirement that descriptions explain WHAT, WHEN, and WHAT IT RETURNS.
14 of 21 tools have NO visible input schema. This prevents LLMs from understanding parameter types, constraints, and required fields. Violates pattern:constrained-input and pattern:tool.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 24 | 2025-06-18+ | v2 |
Forward feeds_catalog to the live membrane.
Forward game_frame_meta to the live membrane.
Forward ide_rebuild_mesh to the live membrane.
Query the math court catalog and grade operating points across multiple domains (geometry, chance, algebra, physics, qcd, health, finance, cs, fluids, eclipse, eht, seismic, disease, chemistry, material, pdb, rife, dynamo, qma_2local, qma_spinglass, qma_nrep, qma_permanent).
Forward membrane_health to the live membrane.
Forward noaa_goes_r_weather to the live membrane.
Route a spin glass manifold problem.
Forward twin.robotics.evaluate_exact_ik to the live membrane.
Forward umc_direct to the live membrane.
Forward umc_resume to the live membrane.
Forward umc_status to the live membrane.
Forward verify_jordan_bond to the live membrane.
Verify n-representability of a density matrix.
Forward weather.convective_containment to the live membrane.
Tool names lack action verbs and use domain jargon (corpus_bonds, critique_frame, umc_direct, ide_rebuild_mesh). Baseline: 90% of A+ tools start with action verbs (get, create, search, update, list). These names provide zero semantic clarity to LLMs about what action occurs. Violates pattern:tool naming requirement.
No output schemas documented for any tool. LLMs cannot plan downstream tool calls or extract required fields. Violates pattern:tool baseline that 100% of A+ tools have documented return types. Agents cannot chain tools without knowing what fields are returned.
No error handling guidance. Tools provide no recovery hints, error classification, or actionable messages. Violates pattern:recovery-guide. When a tool fails, LLMs have no guidance on whether to retry, ask the user, or abort.