Workflow Patterns

When you create a workflow via the wizard, you choose from seven pre-built patterns. Each pattern defines a set of slots — roles that need to be filled by nodes (skills, prompts, services, etc.) to make the workflow functional.

How patterns work

Choose a pattern in the first step of the wizard
Fill the slots — each slot expects a specific node type (skill, prompt, service, source, or asset)
Add custom slots if you need roles beyond what the pattern provides (up to 10 slots total, max 3 per node type)
Review and create — the wizard generates the workflow node and wires up all connections

Required slots must be filled before you can finish. Optional slots can be left empty and added later.

You can always add custom slots beyond the pattern’s template during the Components step of the wizard. This lets you extend any pattern to fit your specific needs.

1. Content Processing

Input → Transform → Output pipeline for converting content between formats.

Use cases: meeting notes to email summary, document reformatting and restructuring, language translation with tone adaptation.

Slot	Node Type	Description	Required	Min	Max
Input Skill	skill	Receives and preprocesses raw input content	Yes	1	1
Transform Skill	skill	Transforms, filters, or processes the content	Yes	1	3
Output Prompt	prompt	Formats and structures the final output	Yes	1	1
Style Guide	source	Reference source for style, tone, or formatting rules	No	0	1

Example: A content team uses this pattern to convert raw interview transcripts into polished blog posts. The input skill cleans and segments the transcript, a transform skill extracts key quotes and themes, and the output prompt formats everything into a publication-ready article. An optional style guide source ensures the output matches the company’s editorial voice.

2. Analysis & Report

Analyse input thoroughly then generate a structured report.

Use cases: code review and quality assessment, security audit and vulnerability report, competitive analysis and market research.

Slot	Node Type	Description	Required	Min	Max
Input Skill	skill	Ingests and prepares the subject for analysis	Yes	1	1
Analysis Skill	skill	Performs the core analysis on the input	Yes	1	2
Report Prompt	prompt	Generates the final report from analysis results	Yes	1	1
Reference Source	source	Standards, checklists, or guidelines for the analysis	No	0	1

Example: A development team uses this pattern for automated code reviews. The input skill parses a pull request diff, the analysis skill checks for security issues, performance concerns, and style violations against a reference source (their coding standards document), and the report prompt produces a structured review with severity ratings and suggested fixes.

3. Classify & Route

Classify input by intent or type, then route to the appropriate handler.

Use cases: customer support triage and response, bug report categorisation and priority assignment, intent-based message routing.

Slot	Node Type	Description	Required	Min	Max
Classification Skill	skill	Analyses input and assigns a category or intent	Yes	1	1
Handler Prompt	prompt	Handles a specific category of classified input	Yes	2	3
Routing Rules	source	Rules or guidelines for routing decisions	No	0	1

Example: A support team uses this pattern to triage incoming tickets. The classification skill reads each ticket and assigns it a category (billing, technical, feature request). Each category has its own handler prompt — one drafts a billing resolution, another pulls up relevant troubleshooting steps, and a third logs feature requests into a backlog format. A routing rules source defines the category definitions and escalation thresholds.

4. Data Extraction

Extract structured data from unstructured input, validate, and format.

Use cases: invoice and receipt processing, form and survey response parsing, log file analysis and metric extraction.

Slot	Node Type	Description	Required	Min	Max
Extraction Skill	skill	Extracts structured fields from unstructured input	Yes	1	1
Validation Skill	skill	Validates and normalises extracted data	Yes	1	1
Output Schema	asset	Schema or template defining the expected output structure	No	0	1

Example: An accounts team uses this pattern to process supplier invoices. The extraction skill pulls out vendor name, invoice number, line items, and totals from PDF text. The validation skill checks for missing fields, normalises date formats, and flags amounts that look unusual. An optional output schema asset defines the exact JSON structure expected by their accounting system.

5. RAG Pipeline

Retrieve relevant context, augment the prompt, and generate an informed response.

Use cases: knowledge base question answering, documentation search and synthesis, research assistant with source citations.

Slot	Node Type	Description	Required	Min	Max
Query Skill	skill	Processes and refines the user query for retrieval	Yes	1	1
Retrieval Service	service	Vector database or search service for context retrieval	Yes	1	1
Augmentation Prompt	prompt	Combines retrieved context with the original query	Yes	1	1
Generation Prompt	prompt	Generates the final response from augmented context	Yes	1	1

Example: A product team builds an internal Q&A tool using this pattern. The query skill rewrites vague user questions into precise search queries. The retrieval service (Pinecone or Chroma) fetches the most relevant documentation chunks. The augmentation prompt stitches the retrieved context into a structured prompt, and the generation prompt produces a clear answer with citations back to the original docs.

6. Multi-Step Generation

Research, draft, review, and polish — ideal for complex content creation.

Use cases: research reports and white papers, long-form writing and blog posts, proposal and pitch document generation.

Slot	Node Type	Description	Required	Min	Max
Research / Context Skill	skill	Gathers background information and context	Yes	1	2
Draft Prompt	prompt	Produces the initial draft from research	Yes	1	1
Review Skill	skill	Reviews the draft for quality, accuracy, and completeness	Yes	1	1
Polish Prompt	prompt	Refines and polishes the final output	Yes	1	1

Example: A consulting firm uses this pattern to generate client proposals. One research skill gathers company background from public sources, another pulls relevant case studies from their internal library. The draft prompt assembles these into a structured proposal. The review skill checks for factual accuracy, completeness, and tone. The polish prompt handles final formatting, executive summary generation, and consistent styling.

7. Human-in-the-Loop

Automated processing with a human review and approval gate.

Use cases: content moderation and compliance review, pull request approval workflows, sensitive communication drafting and sign-off.

Slot	Node Type	Description	Required	Min	Max
Processing Skill	skill	Performs initial automated processing	Yes	1	1
Review Prompt	prompt	Presents results for human review with summary and flags	Yes	1	1
Approval Gate	prompt	Captures the approval decision and any modifications	Yes	1	1
Output Prompt	prompt	Formats the approved output for delivery	Yes	1	1

Example: A legal team uses this pattern for contract review. The processing skill scans a contract and highlights non-standard clauses, missing sections, and risk factors. The review prompt presents a summary with flagged items for a human reviewer. The approval gate captures the reviewer’s decision — approve, reject, or request changes — along with any annotations. Once approved, the output prompt formats the final review memo for the client file.