FigJam Workflow Ingest

Area: IT Workflow: workflows/it/figjam-workflow-ingest.json

Purpose

Reads a FigJam board where the PM places ideas and the bases for a future automation, and returns a well-structured Markdown specification generated by an LLM. The Markdown is the base for engineering to turn the PM’s brainstorm into a concrete workflow.

End-to-end: PM idea on FigJam → structured items → aggregated digest → LLM → workflow spec (Markdown).

Trigger

• Type: Form Trigger.
• Fields:
  • FigJam URL (required) — full board URL.
  • Notes (optional, textarea) — free context for the request.

Nodes

Node	Type	Purpose
Form Trigger: Receive FigJam URL	n8n-nodes-base.formTrigger	Public form entry point.
Code: Extract File Key	n8n-nodes-base.code	Regex-extracts the Figma file key from /board/, /file/, or /design/ URLs.
HTTP Request: Fetch Figma File	n8n-nodes-base.httpRequest	GET https://api.figma.com/v1/files/{file_key} with the httpHeaderAuth credential (“Figma Jhosep”).
Code: Extract Stickies And Text	n8n-nodes-base.code	Walks the document tree, emits one item per sticky / text node.
Code: Aggregate Items	n8n-nodes-base.code	Collapses N extracted items into a single item with a text digest for the LLM.
OpenAI: Generate Workflow Spec	@n8n/n8n-nodes-langchain.openAi	Message a Model (gpt-4o-mini) — turns the digest into a structured Markdown spec.

Per-item shape (from Code: Extract Stickies And Text)

{
  "node_id": "1:23",
  "type": "STICKY | TEXT | SHAPE_WITH_TEXT | STAMP",
  "text": "string",
  "page": "Page name",
  "section": "Section name | null",
  "source_url": "https://www.figma.com/board/...",
  "file_key": "string",
  "extracted_at": "<ISO-8601 timestamp>"
}

Aggregated digest (input to the LLM)

A single item:

{
  "source_url": "...",
  "file_key": "...",
  "extracted_at": "...",
  "item_count": 24,
  "digest": "- [Page A / Section X] (STICKY): ...\n- [Page A] (TEXT): ...\n..."
}

LLM output (final workflow output)

A Markdown document with the following sections, in order:

## Title
## Goal
## Trigger
## Inputs
## Steps
## Outputs
## Integrations / Tools
## Open questions

The LLM groups related stickies, translates non-English content to English, and produces the Markdown only (no preamble, no surrounding code fences).

Status

• active: false — activate per-environment.
• No persistence sink. The Markdown is the workflow output. Persistence (Google Sheets, Notion, file write) is a follow-up decision.

Tags

internal.

Credentials

• Figma: Header Auth (X-Figma-Token: <PAT>). See credentials/README.md.
• OpenAI: openAiApi typed credential.

Open questions

• Persistence sink for the generated Markdown (Notion page, Google Doc, repo file, downstream workflow trigger?).
• Should the LLM also produce a JSON skeleton of the workflow alongside the Markdown?
• Should we capture sticky color / author / position for traceability?
• De-dup strategy if the same FigJam is re-ingested.