Kodexa Metadata Skills
A Claude Code plugin providing skills for authoring Kodexa platform metadata. Each skill includes complete reference schemas, interactive wizards, YAML examples, and common mistake guides.
Skills
| Skill | Description |
|---|
| data-definition | Taxon hierarchies, data types, validation rules, conditional formatting, semantic extraction definitions |
| project-template | Project blueprints — stores, assistants, taxonomies, data forms, status workflows, knowledge sets, triggers, and scheduled jobs |
| module | Model modules (Python ML/AI) and skill modules (file packs) — runtime config, sidecars, inference options, deployment |
| data-form | V2 schema-driven UI forms — UINode component trees, data binding, QuickJS scripting, Bridge API |
| assistant | Event-driven processing pipelines — connections, subscriptions, multi-step orchestration, scheduling |
| task-template | Org-scoped task definitions — options, forms, actions, document family groups, AI naming, team assignment |
| activity-plan | Org-scoped step graphs — CREATE_TASK, EXECUTION, BRIDGE_CALL, SCRIPT, LLM, APPROVAL, AGENT steps with inputs schema and templates |
| knowledge-system | Feature types, item types, knowledge sets — rule-based processing customization with DNF expressions |
| service-bridge | External API proxy endpoints — centralized authentication, header management, caching |
| prompt-template | LLM prompt configurations — extraction prompts, system prompts, knowledge-driven overrides |
| kdx-cli | Kodexa CLI — kubectl-style commands, profile management, sync/deploy workflows, secret management |
Installation
Add the marketplace and install the plugin in Claude Code:
claude plugin marketplace add kodexa-ai/kodexa-metadata-skills
claude plugin install kodexa-metadata-skills
After installation, the skills are automatically available in all Claude Code sessions.
For development/testing, you can also load the plugin directly from a local directory:
claude --plugin-dir /path/to/kodexa-metadata-skills
Usage
Skills are invoked automatically when Claude Code detects a matching task, or you can invoke them explicitly:
# Ask Claude Code to create metadata
> Create a data definition for extracting purchase order data
# Claude Code will automatically use the data-definition skill
# Or invoke directly via the Skill tool
> /kodexa-metadata-skills:data-definition
Interactive Wizard
Each skill offers a guided wizard mode. When you ask to create metadata from scratch, the skill walks you through:
- Purpose and use case questions
- Field and component selection
- Validation and business rule configuration
- Complete YAML generation
Reference Mode
When editing existing metadata or debugging issues, the skill loads the complete schema reference into context to assist with any modification.
Skills Detail
Tier 1 — Core
data-definition — Author taxonomy YAML for document extraction. Covers all 12 data types (STRING, NUMBER, CURRENCY, DATE, SELECTION, etc.), value paths, validation formula functions, conditional formatting, repeating groups, and semantic definitions.
project-template — Author project blueprint YAML. Covers stores, assistants, taxonomies, data forms, document/task/attribute status workflows, inline and org-level task templates, scheduled jobs, knowledge sets, project options (with companion + execution policy), and how triggers and activity-plans bind to projects.
module — Author module YAML for both model (Python) and skill (file pack) types. Covers runtime configuration, sidecar dependencies, inference options, deployment defaults, Python entry point patterns, and magic parameter injection.
Tier 2 — Project Components
data-form — Author V2 schema-driven forms in JSON. Covers UINode component tree, layout/display/data components, Bridge API namespaces (kodexa.data, kodexa.navigation, kodexa.form, kodexa.http, kodexa.log), event handling, and scripting.
assistant — Author assistant pipeline YAML. Covers connection types (STORE, CHANNEL, DOCUMENT_FAMILY, WORKSPACE), subscription expressions, multi-step pipeline chaining, extension pack structure, and Python implementation.
task-template — Author org-scoped task template YAML. Covers metadata.options (unified field/option list with groupOptions and listType), forms, action buttons, document family groups, AI naming, chat prompt, execution policy, companion config, and the initialStatusSlug that resolves against project-bound task-statuses.
