Loading ToolBench
Nuke adapter for the DCC Model Context Protocol ecosystem
Server provides 10 tools with basic schemas and descriptions, but quality is inconsistent. Tool names follow verb_noun convention (create_node, delete_node, get_knob, set_knob, list_node_graph). Descriptions exist but are terse (avg ~60 chars, below 194-char baseline). Critical gap: parameter descriptions are minimal or missing context. The 'nuke' parameter (object type) appears in 8 tools but lacks meaningful description beyond 'Nuke module instance'. Input schemas present but lack depth, no enums, ranges, or validation constraints. Output schemas are not documented. Error handling is not visible in tool definitions. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications in metadata.
Replace the current graph with a layered Read/Merge/Write composite
Connect or disconnect one input, rolling back if readback disagrees
Create a Gizmo from a Group node with exposed knobs and input contract
Create one available node class without clearing or replacing existing nodes
Delete one exact node and verify that it is gone
Read one non-executable knob as a bounded JSON value
Return bounded node identity, topology, and JSON-safe knob values
Parameter descriptions are generic or missing. The 'nuke' parameter (object type) appears in 8 tools with only 'Nuke module instance', no guidance on what properties it exposes or how to obtain it. LLMs cannot infer parameter meaning from names alone.
No input validation constraints (enums, ranges, patterns) visible in schemas. E.g., 'input_index' (0-63) lacks min/max bounds; 'conflict_policy' lacks enum values; 'max_nodes' and 'max_knobs_per_node' lack documented ranges despite description mentioning '1-1000' and '0-128'.
Output schemas are not documented. Tools like 'list_node_graph' and 'validate_manifest' return complex structures, but no schema is provided for what fields the LLM should expect. This forces LLMs to guess at response structure.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
Readiness diagnostic tool for Nuke host
Write one static non-executable knob and require a matching readback
Validate and normalize a deterministic layered-composite manifest
Tool annotations missing. Despite clear risk classifications (READ_ONLY, WRITE, DESTRUCTIVE), no readOnlyHint, destructiveHint, or idempotentHint annotations are present in tool definitions. Agents cannot distinguish safe from dangerous operations.
Error handling and recovery guidance not visible. No indication of what errors tools can raise, how to classify them (retryable vs fatal), or what the LLM should do next. E.g., 'delete_node' could fail if node doesn't exist, no guidance provided.