Add workflow orchestrator skill with schema and templates

[GITDIDDY69]

Summary

Introduces the workflow-orchestrator skill, enabling users to define and execute multi-step workflows as directed acyclic graphs (DAGs) with dependency tracking, parallel execution, and failure handling.

Key Changes

• SKILL.md: Complete skill documentation covering:

  • Core workflow concepts (steps, dependencies, parallel execution)
  • Workflow definition in both YAML and natural language formats
  • State tracking via workflow-state.json for progress and recovery
  • Execution algorithm with wave-based dispatch, retry/fallback handling, and output chaining
  • Agent dispatch pattern for step invocation
  • Resume capability for interrupted workflows

• workflow-schema.md: Comprehensive JSON Schema reference including:

  • Workflow object structure with name, description, steps, and global failure/timeout settings
  • Step object with id, skill/tool, input, dependencies, retry, fallback, and conditional execution
  • Retry configuration (max_attempts, modified_input with error references)
  • Fallback configuration for alternative execution paths
  • State file schema tracking workflow and per-step execution status
  • Template reference format for parameterized workflows
  • Validation rules (unique IDs, valid references, acyclic dependencies, template resolution)

• templates.md: Four predefined workflow templates for common patterns:

  • Research and Report: Multi-source research with synthesis and document generation
  • Build, Test, Deploy: Code quality pipeline with linting, testing, and deployment
  • Data Pipeline: ETL workflow with extraction, profiling, transformation, validation, and loading
  • Content Creation: Research-to-output workflow with drafting, editing, and multi-format production

Each template includes parameter definitions, step-by-step workflow structure, and customization suggestions.

Implementation Details

• Workflows are DAGs with steps executed in dependency-respecting waves, enabling parallel execution of independent steps
• Template references ({{step_id}}) allow downstream steps to consume upstream outputs
• State file enables workflow resumption and progress tracking across interruptions
• Flexible failure handling: abort, skip, fallback, or ask user
• Supports both structured (YAML) and natural language workflow definitions with automatic parsing and user confirmation

https://claude.ai/code/session_01HP4LKYY14rwNjEtd7v71ut