{{ outputs.scan_repo | truncate(2000) }} ---

      Write the source to: `{{ inputs.workspace_dir }}/diagrams/arch.puml`
      (create parent dir if missing; overwrite OK).

      Reply with the absolute output path on a single line, no preamble.
- id: render_drawio
  kind: agent
  skill: sub-agent
  depends_on: [scan_repo, classify_kind]
  with:
    task: |
      Generate a draw.io XML diagram of kind `{{ outputs.classify_kind }}`
      from this codebase scan. Use valid draw.io XML:
        <mxfile><diagram><mxGraphModel><root>
          <mxCell id="0"/><mxCell id="1" parent="0"/>
          <mxCell id="N" value="..." style="..." vertex="1" parent="1">
            <mxGeometry .../>
          </mxCell>
          <mxCell ... edge="1" source="..." target="..." parent="1">
            <mxGeometry .../>
          </mxCell>
        </root></mxGraphModel></diagram></mxfile>

      Aim for 10-20 boxes/edges to mirror the PlantUML side; layout can be
      simple grid since the user is expected to re-arrange in draw.io.

      Scan output:
      ---
      {{ outputs.scan_repo | truncate(2000) }}
      ---

      Write the XML to: `{{ inputs.workspace_dir }}/diagrams/arch.drawio`
      (create parent dir if missing; overwrite OK).

      Reply with the absolute output path on a single line, no preamble.
- id: compose_doc
  kind: agent
  skill: docx
  depends_on: [render_plantuml, render_drawio]
  with:
    task: |
      Compose an architecture document that references both diagram
      deliverables. The doc body should contain three sections:

      1. **Scope** — restate the user's invocation target.
         User invocation: {{ inputs.user_message | xml_escape | truncate(200) }}
      2. **Diagram kind** — `{{ outputs.classify_kind }}` (one paragraph
         justifying why this kind was chosen).
      3. **Deliverables** — two bullet items linking the files:
         - PlantUML source: {{ outputs.render_plantuml }}
         - draw.io file: {{ outputs.render_drawio }}

      Save to: `{{ inputs.workspace_dir }}/diagrams/arch.docx`. Reply
      with the absolute output path on a single line, no preamble.
- id: persist
  kind: tool_call
  tool: memory_save
  tool_allowlist: [memory_save]
  depends_on: [compose_doc]
  tool_args:
    path: "memory/architecture-snapshots.md"
    mode: append
    content: |
      === diagram triangulation ===
      invocation: {{ inputs.user_message | xml_escape | truncate(200) }}
      diagram_kind: {{ outputs.classify_kind }}
      plantuml: {{ outputs.render_plantuml | truncate(200) }}
      drawio: {{ outputs.render_drawio | truncate(200) }}
      docx: {{ outputs.compose_doc | truncate(200) }}

Diagram Triangulation (Meta-Skill)

A classifier + parallel render meta-skill. After scanning a target path, an llm_classify step picks the most informative diagram kind (one of class | sequence | component | deploy | flow), then two independent render branches synthesize PlantUML source and draw.io XML for the same scan — running in parallel because they are independent tools serving different downstream uses (git-friendly text review vs. editable canvas).

Trigger surface

Fire by saying diagram triangulation or one of the localized triggers listed in the frontmatter, with the target path or module reference in the same turn.

Fallback

If either render step fails, compose_doc should still produce the docx referencing whichever render succeeded; manually re-run the failed render via sub-agent with the same scan output as input.