Workflow Inputs and Step References in PromptPlan

Supported reference syntax

Workflow step prompts support three core references:

{{workflow.input.KEY}}
{{step.previous.output}}
{{step.STEP_KEY.output}}

Workflow input references

Use workflow inputs when a value should be filled once at run time and reused across multiple steps.

Example:

Write for {{workflow.input.audience}} in a {{workflow.input.tone}} tone.

Best use cases:

• audience
• tone
• source material
• feature brief
• constraints

Previous-step references

Use {{step.previous.output}} when the step should consume the output from the immediately previous enabled step.

Example:

Review the draft below and improve structure:

{{step.previous.output}}

This keeps linear workflows simple and readable.

Named step references

Use {{step.STEP_KEY.output}} when a step needs a specific earlier artifact instead of the most recent one.

Example:

Use the summary and the key points together:

Summary:
{{step.summarize_article.output}}

Key points:
{{step.extract_key_points.output}}

Input type guidance

Choose input types deliberately:

• use text for short labels and short instructions
• use long_text for source documents, briefs, notes, or transcripts
• use select when you want controlled choices like tone or target platform
• use boolean for simple toggles
• use json only when you actually need structured payloads

Common mistakes

Avoid these patterns:

• putting static instructions into inputs instead of the prompt template
• using step.previous.output when a later step really depends on a named earlier artifact
• naming step keys casually and then changing them often
• storing giant multi-purpose payloads in one input when two or three smaller inputs would be clearer

Recommended naming style

Use stable, descriptive keys:

• feature_description
• target_platform
• interview_notes
• extract_pain_points
• export_brief

This keeps prompt templates readable and safer to edit later.