PluginBench
Skill

deepline-plays

code.deepline.com

Create custom Deepline plays combining multiple tools, datasets, and fallback logic with advanced run/export behavior.

What is deepline-plays?

Deepline Plays lets you build reusable scripts that orchestrate multiple tools and other plays together, with support for durable datasets, fallback logic, joins, projections, and custom execution behavior. Use this when you need to compose complex workflows that go beyond single-tool operations.

  • Combine multiple tools and other plays into a single reusable script
  • Define durable datasets that persist across play executions
  • Implement fallback logic for robust error handling
  • Join and project data across multiple sources
  • Customize run and export behavior for specific workflows
  • Orchestrate complex multi-step operations

How to install deepline-plays

npx skills add null --skill deepline-plays
Prerequisites
  • Deepline environment access
  • Understanding of deepline-gtm meta-skill (which governs the session)
  • Access to the deepline-plays recipe file at ../deepline-gtm/recipes/deepline-plays.md
Claude Code
Cursor
Windsurf
Cline

How to use deepline-plays

  1. 1.Invoke the deepline-gtm meta-skill using the Skill tool
  2. 2.Follow the meta-skill's full routing instructions and load all sub-documents it specifies
  3. 3.Read the deepline-plays recipe at ../deepline-gtm/recipes/deepline-plays.md for the specific workflow
  4. 4.Define your play structure combining tools and/or other plays
  5. 5.Configure datasets, fallback logic, joins, and projections as needed
  6. 6.Customize run and export behavior for your use case
  7. 7.Test and deploy your play

Use cases

Good for
  • Building reusable workflow templates that chain multiple Deepline tools together
  • Creating data pipelines that combine tool outputs with fallback strategies
  • Designing parameterized scripts that teams can invoke repeatedly with different inputs
  • Implementing complex data transformations using joins and projections
  • Automating multi-step processes with custom execution and export logic
Who it's for
  • Deepline users building reusable automation workflows
  • Teams creating standardized process templates
  • Data engineers designing multi-step pipelines
  • Developers building complex tool orchestration scripts

deepline-plays FAQ

How does deepline-plays relate to deepline-gtm?

deepline-plays is a recipe shortcut that pre-selects the plays workflow, but deepline-gtm is the meta-skill that governs the entire session. You must invoke deepline-gtm and follow its full routing instructions; the recipe only covers one part of the task.

Can I combine multiple plays together?

Yes, plays can reference and combine other plays along with individual tools, allowing you to build hierarchical and reusable workflow compositions.

What happens if a tool fails in my play?

You can implement fallback logic to handle failures gracefully, ensuring your play continues or switches to an alternative approach when a tool encounters an error.

Can I customize how my play exports data?

Yes, deepline-plays supports custom run and export behavior, allowing you to control how results are formatted and delivered.

Full instructions (SKILL.md)

Source of truth, from code.deepline.com.


name: deepline-plays description: "Create custom Deepline plays/scripts that combine multiple tools and/or other plays, with durable datasets, fallback logic, joins, projections, and custom run/export behavior."

Deepline Plays Recipe

This is a recipe shortcut. It pre-selects the deepline-plays recipe but the deepline-gtm governs the entire session.

Execution order

  1. Invoke deepline-gtm using the Skill tool.
  2. Follow the meta-skill's full routing instructions - analyze the user's complete prompt and load every sub-doc the meta-skill tells you to. Do not skip docs just because a recipe is pre-selected.
  3. Additionally read the deepline-plays recipe at ../deepline-gtm/recipes/deepline-plays.md (relative to this file) for the specific workflow.

The recipe only covers one part of the task. The meta-skill handles everything else the user asked for.