Loading ToolBench

Reading the current corpus…

datahub-steward-squad MCP server · D · ToolBench

datahub-steward-squad

github.com/maharshi-coding/datahub-steward-squad

ECC-style multi-agent steward team for DataHub metadata governance. Runs a DataHub-shaped MCP server (JSON-RPC 2.0 over stdio) that serves metadata through read tools (search, get_entities, get_lineage, list_schema_fields) and optional mutation tools (update_description, add_tags, add_terms, save_document).

2026-07-28+STDIO8 tools

54/100

Definition Quality66

Protocol Readiness50

Supportability38

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence

Current-spec patterns detected

Summary

Server has 8 tools with complete input schemas and basic descriptions, but significant gaps in parameter documentation and output schema clarity. All tools have descriptions (10-100 chars), but 5 of 8 lack parameter-level descriptions, critical for LLM reasoning. No output schemas documented. Mutation tools (update_description, add_tags, add_terms, save_document) lack error handling guidance. Read tools are well-named (search, get_entities, get_lineage, list_schema_fields) but parameter descriptions are minimal or missing. No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Composition is sound, tools are single-responsibility and chainable, but parameter relationships undocumented.

Tools (8)

add_tagswritesource verified65/100

Attach tags to an entity or one of its schema fields.

add_termswritesource verified65/100

Attach glossary terms to an entity or one of its schema fields.

get_entitiesread onlysource verified72/100

Fetch full metadata (schema fields, assertions, owners) for URNs.

get_lineageread onlysource verified70/100

Return upstream/downstream lineage edges for a URN.

list_schema_fieldsread onlysource verified68/100

List schema fields with tags and glossary terms for a dataset URN.

save_documentwritesource verified60/100

Save a context document into DataHub.

searchread onlysource verified68/100

Search DataHub assets by free-text query, domain, or entity type.

update_description

Issues (5)

CRITICALtool-description ↗

Parameter descriptions missing or minimal. 'query', 'domain', 'entity_type' in search lack descriptions. 'mode' in update_description has no description. LLMs cannot infer parameter semantics from names alone.

searchupdate_descriptionadd_tagsadd_termssave_document

HIGHresponse-shaper ↗

No output schemas documented. Tools return results but LLMs have no formal specification of response structure, field types, or chaining IDs. Agents cannot plan downstream calls reliably.

searchget_entitiesget_lineagelist_schema_fields

HIGHrecovery-guide ↗

Mutation tools lack error handling guidance. No recovery hints for failed updates, permission errors, or invalid URNs. Agents cannot self-correct on failure.

update_descriptionadd_tagsadd_termssave_document

Recommendations

Add parameter-level descriptions to all 8 tools. For 'query' in search: 'Free-text search string (e.g., "customer_data", "revenue_model")'. For 'mode' in update_description: 'Update mode: "replace" (default) or "append"'. For 'field_path' in add_tags/add_terms: 'Optional schema field path (e.g., "customer_id"). If omitted, tags/terms apply to the entity itself.'
Document output schemas for all tools. E.g., search returns [{urn, name, entity_type, domain, description}]. get_entities returns {urn, name, description, schema_fields: [{name, type, tags, terms}], owners, assertions}. Include field types and whether fields are always present.
Add tool annotations: readOnlyHint=true for search, get_entities, get_lineage, list_schema_fields; destructiveHint=true for update_description, add_tags, add_terms, save_document (if they overwrite). This enables agents to reason about side effects.
Add error handling guidance to mutation tools. E.g., update_description: 'Returns error if URN not found (try search() first) or if user lacks write permission. Errors include the invalid URN and available entity types.'
Specify constraints on numeric/enum parameters. E.g., get_lineage depth: 'Integer 1 - 10 (default 2). Deeper traversals may timeout.' search entity_type: 'One of: dataset, dashboard, mlmodel, datajob, domain, glossary_term. Omit to search all types.'

Score history

Overall score trend

First recorded score · v2 rubric

54/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	D	54	2026-07-28+	v2