Imported from LoxtepInc/loxtep-plugins-skills (opencode/skills/loxtep-semantic-layer/SKILL.md). Install upstream with npx skills add LoxtepInc/loxtep-plugins-skills --skill loxtep-semantic-layer. Copyright stays with the author.

name: loxtep-semantic-layer description: Use when the user wants to search the semantic layer, retrieve semantic artifacts, check semantic completeness, or manage canonical knowledge (strategy, positioning, brand voice, org structure, process docs). metadata: documentation: https://github.com/LoxtepInc/loxtep-plugins-skills/blob/main/opencode/skills/loxtep-semantic-layer/SKILL.md

Semantic Layer (Customer MCP)

Search and inspect the organization's semantic layer - curated business definitions, metrics, dimensions, and completeness scores. Also manage canonical knowledge artifacts like strategy, positioning, brand voice, org structure, and process documentation.

When to use

• "Search semantic layer", "find semantic artifact", "completeness score"
• "What metrics / dimensions / definitions exist for X?"
• "How complete is the semantic layer?"
• "Upload unstructured context knowledge", "add canonical knowledge", "create/update strategy", "brand voice", "org structure", "process documentation"
• "Store business strategy for agents", "Define our brand positioning", "Document organizational structure"

Prerequisites

• MCP auth. Operations are organization-scoped.

Happy-path flow

Search and retrieval

1. search_semantic_layer with a query term to discover relevant artifacts.
2. get_semantic_artifact by artifact_type + id to inspect a specific artifact.
3. get_semantic_completeness to assess overall semantic coverage.

Canonical knowledge management

1. create_canonical_knowledge to store organizational knowledge (strategy, positioning, brand voice, org structure, process docs).
2. get_canonical_knowledge to retrieve specific knowledge artifacts.
3. update_canonical_knowledge to modify existing knowledge bases.

Operations

MCP mapping

Canonical Knowledge Types

The semantic layer supports structured organizational knowledge artifacts that agents can query for context:

Creating Canonical Knowledge

{
  "operation": "create_canonical_knowledge",
  "type": "strategy",
  "title": "Q4 2024 Revenue Strategy",
  "body": "Our strategy is to increase ARR by 40% through expansion of enterprise accounts...",
  "classification": "internal",
  "owner": "strategy-team@company.com"
}

Retrieving Knowledge

{
  "operation": "get_canonical_knowledge",
  "id": "<knowledge-id>"
}

Updating Knowledge

{
  "operation": "update_canonical_knowledge",
  "id": "<knowledge-id>",
  "title": "Updated Q4 2024 Revenue Strategy",
  "body": "We've adjusted our targets to 45% ARR growth..."
}

Pitfalls

• Ontology/vocabulary management (thesaurus terms, namespace mappings) is loxtep_meaning - different facade. This Agent-Scope Skill is for querying the curated semantic layer and managing canonical knowledge, not managing its underlying ontology structure.
• search_semantic_layer with artifact_types: schema is not domain schemas. It searches schema_registry_cache / curated artifacts — not domain_schemas from create_schema. Org shapes live under loxtep_define list_schemas. Pack hits for “Product” are concepts, not the S&S Product shape you applied to a raw data product.
• Do not bind a concept onto a DP that already has an applied shape. Apply the shape to the product; align the shape to the concept in Meaning. See docs/concepts/type-vs-pack-alignment.md.
• Catalog search is loxtep_query - use that for broad discovery across all artifact types. This Agent-Scope Skill is for the semantic-layer-specific search and completeness view.
• artifact_types (plural) is an array filter - pass multiple types to narrow results. Do not confuse with artifact_type (singular) used in get_semantic_artifact.
• Canonical knowledge is for organizational knowledge, not data product schemas or quality rules. Use loxtep_define for schema management and loxtep_define for quality rules.
• Do NOT use workflows, data products, or connectors for storing unstructured organizational knowledge. Use canonical knowledge artifacts instead.

COA list_metrics / describe_schema gaps

There are no hosted MCP ops named list_metrics or describe_schema. Use:

Agent-Scope Skill scope (.loxtep/skills/loxtep-semantic-layer.yaml)

Resource scope and operation permissions for this Agent-Scope Skill, conformant with the skill-package-v1 schema. Any resource type or operation not listed is denied (fail-closed). Identifier lists are empty placeholders — fill them with the specific resources in your workspace. This declaration does not change the hosted MCP config (mcp.loxtep.io).

# .loxtep/skills/loxtep-semantic-layer.yaml
# Conforms to https://loxtep.io/schemas/skill-package-v1.json
# Fail-closed: this Agent-Scope Skill's facades are RBAC-governed and carry no data-mesh resource scope.
name: loxtep-semantic-layer
description: Semantic layer search, artifact retrieval, and completeness — RBAC-governed; no data-mesh resource scope.
scope:
  data_products: []
  connectors: []
  workflows: []
  domains: []
  queues: []
permissions: {}

Optional attribution

_metadata: { "skill_name": "loxtep-semantic-layer" }

Auth

loxtep-auth / login.

References

• User story catalog
• COA MCP serve parity