Loading ToolBench
Typed domain plugin for developing and maintaining Yandex DataLens objects through Codex and MCP clients.
Server provides 23 tools with complete input schemas and descriptions. Naming follows verb_noun convention (dl_* prefix). However, descriptions are often generic and lack actionable context for LLM selection. Output schemas are not documented. Error handling guidance is absent. Parameter descriptions exist but lack format constraints, ranges, and dependency documentation. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Composition is reasonable but some tools combine multiple concerns (e.g., dl_object_update handles patches, artifacts, and dashboard-specific logic).
Returns administrative capabilities including license management, workbook lifecycle, and HTML page support.
Checks authentication status and credential configuration. Returns credential report and health status.
Refreshes authentication credentials and probes their validity. Optionally allows browser-based authentication.
Retrieves authoring defaults for visual profiles including project root, family, and presentation settings.
Creates a backup of DataLens objects for recovery purposes.
Identifies and optionally removes unused DataLens objects.
Output schemas not documented. Tools return complex nested structures (dashboards, datasets, revisions) but LLMs have no schema to understand response fields, forcing unstructured parsing.
Descriptions lack actionable context. Generic phrases like 'Retrieves...' and 'Lists...' do not explain WHEN to use each tool or how it differs from similar tools (e.g., dl_object_get vs dl_dashboard_snapshot both retrieve objects).
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Tools marked DESTRUCTIVE (dl_object_delete) and WRITE (dl_object_update, dl_object_create) should carry annotations to guide agent behavior.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2025-06-18+ | v2 |
Compiles a recipe into DataLens objects with bindings and optional presentation customization.
Retrieves a complete snapshot of a dashboard including all tabs, items, and connections.
Previews dataset data with filtering, sorting, and parameter support.
Validates dataset fields and optionally performs provider-level validation with source refresh.
Retrieves the schema for a specific DataLens API method from the operation registry.
Creates a new DataLens object with specified type and initial content.
Deletes a DataLens object permanently.
Retrieves a DataLens object (dashboard, dataset, chart, etc.) with optional field filtering and view selection.
Publishes a saved object revision to the published branch.
Retrieves relationships between objects with pagination support.
Lists revision history for an object with optional filtering by revision IDs.
Updates an existing DataLens object with changes and optional dashboard-specific patches.
Returns server information including name, version, architecture, and runtime identity.
Validates draft objects (dashboards, charts, datasets) for correctness and completeness.
Lists entries (objects) within a specific workbook with pagination support.
Lists all workbooks with pagination support.
Validates HTML content for size and format compliance.
Parameter descriptions lack format constraints and ranges. E.g., page_size defaults to 100 but no min/max specified; branch enum is clear but other string params (object_id, method) lack validation hints.
Error handling guidance absent. No recovery hints in tool descriptions. E.g., dl_object_get with invalid object_id should suggest 'Try dl_workbooks_list() to discover valid IDs' but does not.