Loading ToolBench
Model Context Protocol (MCP 2026-07-28) Universal Adapter for Conductor. Exposes canonical NEURONIX Skills as MCP tools and Vital observations as MCP resources. Adheres strictly to SPEC-NRX-CND-018, SPEC-NRX-SKL-019, and SPEC-NRX-CND-021.
Server exposes 9 tools with schemas and descriptions, but critical gaps undermine production readiness. Tool names lack consistent verb-noun structure (neuronix_* prefix obscures intent). Descriptions are present but generic, most under 100 chars, lacking WHEN/WHY context. Input schemas exist but many parameters lack descriptions (e.g., neuronix_diet's dry_run has no description; neuronix_shadow_eval's config_path is vague). No output schemas documented. Error handling is minimal, no recovery guidance, no categorization of retryable vs fatal errors. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Composition is reasonable (each tool has one responsibility), but parameter naming is inconsistent (dry_run vs mode enum). Security: no evidence of secret injection, permission gates, or audit logging in visible code.
Check upstream flake repository and remote commit status for available system updates.
Trigger storage garbage collection, inode hardlink deduplication, and VirtIO TRIM unmap directives to shrink host disk.
Execute deep system diagnostic probe and produce privacy-sanitized system report for issue tracking.
List historical system generations with active generation marker and symlink targets.
Simulate system configuration in an ephemeral in-memory Shadow Micro-VM in RAM (/dev/shm) with automated smoke test before host promotion.
Inspect OS kernel parameters, active NixOS generation, storage telemetry, and autonomous timer status.
Tool names lack action verbs. 'neuronix_status', 'neuronix_diet', 'neuronix_shadow_eval' obscure intent. LLMs cannot infer whether these read, write, or modify state from names alone. Should be 'get_system_status', 'trigger_storage_cleanup', 'simulate_config_in_sandbox'.
Parameter descriptions missing or vague. neuronix_diet.dry_run has no description. neuronix_shadow_eval.config_path says 'Optional path' but does not explain what happens if omitted or what format is expected. neuronix_upgrade.mode enum values lack descriptions.
No output schemas documented. Tools return results but LLMs cannot plan downstream calls without knowing field names and types. E.g., neuronix_list_generations returns 'historical system generations' but structure is unknown.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2026-07-28+ | v2 |
Execute atomic system rollback to preceding generation (switches generation when elevated, or returns exact recovery directive).
Perform atomic system upgrade with generation creation (defaults to staged mode to prevent active session disruption).
Evaluate and validate a Nix package or expression derivation via dry-build dependency resolution prior to execution (Declarative Build Verification Gatekeeper).
No tool annotations despite clear risk levels. neuronix_diet (WRITE), neuronix_undo (REVERSIBLE), neuronix_upgrade (WRITE) lack destructiveHint/idempotentHint. Agents cannot distinguish safe from risky operations.
Error handling absent. No recovery guidance, no categorization of retryable vs fatal errors, no actionable error messages. If neuronix_verify fails, LLM has no guidance on next steps.