Corvex exhibits significant definition quality gaps. While 16 tools are registered with names and descriptions, most descriptions are generic and lack actionable context for LLM selection. Parameter schemas are present but sparse: many lack type constraints, enums, or validation guidance. No output schemas are documented. Error handling is absent, tools provide no recovery guidance. Naming is inconsistent (run_analysis, run_optimization, run_dft_calculation, run_molecular_dynamics, run_task all use 'run_' prefix, creating ambiguity). Critical parameters like 'skill' accept both string and array but lack enum constraints or format guidance. No tool declares permissions, audit trails, or idempotency guarantees. The server is domain-specialized (materials science) but does not leverage that to provide rich, LLM-optimized descriptions.
Generate an initial experimental plan with optional skill constraints and best-of-N planning
Generate crystal structures from text descriptions using LLM-guided structure generation
Retrieve bounded action history from the current session
List all saved literature searches in the current campaign with section references
Generate Python code to prepare and structure experimental inputs for analysis
Extract plain text from PDF, DOCX, Markdown, or text files with optional OCR for scanned pages
No output schemas documented. Tools return unstructured results; LLMs cannot plan downstream calls or extract typed fields. Violates pattern:response-shaper.
Ambiguous naming: 5 tools use 'run_*' prefix (run_analysis, run_optimization, run_dft_calculation, run_molecular_dynamics, run_task). LLMs cannot distinguish intent. Violates pattern:tool naming rule.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | 2026-07-28+ | v2 |
Refine an existing structure with targeted modifications
Execute analysis on experimental data with optional skill selection and best-of-N candidate generation
Execute periodic DFT calculation on a structure
Execute molecular dynamics simulation on a structure
Execute Bayesian optimization with optional skill guidance and surrogate/acquisition selection
Delegate a task to a specialist orchestrator (analysis, planning, or simulation)
Search scientific literature with optional query optimization and multi-source aggregation
Get current session status including generated structures, calculation parameters, and routing decisions
Upload and probe experimental data files for content-based routing
Validate a structure for physical and chemical feasibility
Parameter 'skill' accepts string|array but lacks enum constraints, format guidance, or examples. LLMs cannot infer valid skill names or when to pass array vs string. Violates pattern:constrained-input.
No error handling guidance. Tools provide no recovery hints, retryability classification, or actionable error messages. Violates pattern:recovery-guide.
No permission declarations, audit trails, or idempotency guarantees. Destructive tools (generate_structure, run_dft_calculation, run_molecular_dynamics, upload_files) lack confirmation patterns. Violates pattern:permission-gate and pattern:confirmation-request.