{
    "kind": "meta",
    "name": "meta-diagram-triangulation",
    "always": false,
    "triggers": [
        "diagram triangulation",
        "triangulate diagrams",
        "架构三视图",
        "出双视图架构图"
    ],
    "provenance": {
        "origin": "opensquilla-original",
        "license": "Apache-2.0"
    },
    "composition": {
        "steps": [
            {
                "id": "scan_repo",
                "kind": "agent",
                "with": {
                    "task": "Scan the target path identified in the user's invocation for\narchitectural structure. Extract:\n  * Top-level modules \/ packages under the target path\n  * Inter-module import dependencies (who imports who)\n  * Hotspot files (most-changed in the last 90 days)\n  * Public surface: classes, functions, protocols exported via __init__.py\n\nUser invocation (target path is somewhere in this string):\n{{ inputs.user_message | xml_escape | truncate(400) }}\n\nReply with a structured summary, max 1500 chars:\n  ## Target path\n  <resolved absolute path>\n\n  ## Modules (top-level)\n  - <module\/>: <one-line description>\n\n  ## Dependencies\n  <module> → <module>\n  ...\n\n  ## Hotspots\n  - <file>: <commit count last 90d>\n\n  ## Public surface\n  - <ClassName>: <one-line role>\n"
                },
                "skill": "history-explorer"
            },
            {
                "id": "classify_kind",
                "kind": "llm_classify",
                "with": {
                    "task": "Based on this codebase scan, which diagram kind is most informative\nfor an architecture doc? Pick ONE of:\n  - class: data structures \/ OO hierarchy dominant\n  - sequence: cross-module call flows dominant\n  - component: module-level boxes + arrows dominant\n  - deploy: infra \/ process layout dominant\n  - flow: data pipeline \/ staged processing dominant\n\nScan output:\n{{ outputs.scan_repo | truncate(1500) }}\n"
                },
                "depends_on": [
                    "scan_repo"
                ],
                "output_choices": [
                    "class",
                    "sequence",
                    "component",
                    "deploy",
                    "flow"
                ]
            },
            {
                "id": "render_plantuml",
                "kind": "agent",
                "with": {
                    "task": "Generate a PlantUML diagram source of kind `{{ outputs.classify_kind }}`\nfrom this codebase scan. Use idiomatic PlantUML syntax bracketed by\n`@startuml` ... `@enduml`. Aim for 10-20 boxes\/arrows; do not over-render.\n\nScan output:\n---\n{{ outputs.scan_repo | truncate(2000) }}\n---\n\nWrite the source to: `{{ inputs.workspace_dir }}\/diagrams\/arch.puml`\n(create parent dir if missing; overwrite OK).\n\nReply with the absolute output path on a single line, no preamble.\n"
                },
                "skill": "sub-agent",
                "depends_on": [
                    "scan_repo",
                    "classify_kind"
                ]
            },
            {
                "id": "render_drawio",
                "kind": "agent",
                "with": {
                    "task": "Generate a draw.io XML diagram of kind `{{ outputs.classify_kind }}`\nfrom this codebase scan. Use valid draw.io XML:\n  <mxfile><diagram><mxGraphModel><root>\n    <mxCell id=\"0\"\/><mxCell id=\"1\" parent=\"0\"\/>\n    <mxCell id=\"N\" value=\"...\" style=\"...\" vertex=\"1\" parent=\"1\">\n      <mxGeometry ...\/>\n    <\/mxCell>\n    <mxCell ... edge=\"1\" source=\"...\" target=\"...\" parent=\"1\">\n      <mxGeometry ...\/>\n    <\/mxCell>\n  <\/root><\/mxGraphModel><\/diagram><\/mxfile>\n\nAim for 10-20 boxes\/edges to mirror the PlantUML side; layout can be\nsimple grid since the user is expected to re-arrange in draw.io.\n\nScan output:\n---\n{{ outputs.scan_repo | truncate(2000) }}\n---\n\nWrite the XML to: `{{ inputs.workspace_dir }}\/diagrams\/arch.drawio`\n(create parent dir if missing; overwrite OK).\n\nReply with the absolute output path on a single line, no preamble.\n"
                },
                "skill": "sub-agent",
                "depends_on": [
                    "scan_repo",
                    "classify_kind"
                ]
            },
            {
                "id": "compose_doc",
                "kind": "agent",
                "with": {
                    "task": "Compose an architecture document that references both diagram\ndeliverables. The doc body should contain three sections:\n\n1. **Scope** — restate the user's invocation target.\n   User invocation: {{ inputs.user_message | xml_escape | truncate(200) }}\n2. **Diagram kind** — `{{ outputs.classify_kind }}` (one paragraph\n   justifying why this kind was chosen).\n3. **Deliverables** — two bullet items linking the files:\n   - PlantUML source: {{ outputs.render_plantuml }}\n   - draw.io file: {{ outputs.render_drawio }}\n\nSave to: `{{ inputs.workspace_dir }}\/diagrams\/arch.docx`. Reply\nwith the absolute output path on a single line, no preamble."
                },
                "skill": "docx",
                "depends_on": [
                    "render_plantuml",
                    "render_drawio"
                ]
            },
            {
                "id": "persist",
                "kind": "tool_call",
                "tool": "memory_save",
                "tool_args": {
                    "mode": "append",
                    "path": "memory\/architecture-snapshots.md",
                    "content": "=== diagram triangulation ===\ninvocation: {{ inputs.user_message | xml_escape | truncate(200) }}\ndiagram_kind: {{ outputs.classify_kind }}\nplantuml: {{ outputs.render_plantuml | truncate(200) }}\ndrawio: {{ outputs.render_drawio | truncate(200) }}\ndocx: {{ outputs.compose_doc | truncate(200) }}"
                },
                "depends_on": [
                    "compose_doc"
                ],
                "tool_allowlist": [
                    "memory_save"
                ]
            }
        ]
    },
    "description": "Scan a target codebase path, classify the most informative diagram kind, then render it as BOTH a PlantUML source file AND a draw.io XML in parallel, and compose them into a single architecture doc. Use when writing an RFC or onboarding doc and you want a text-friendly (PlantUML) and an editable (drawio) view of the same architecture.",
    "meta_priority": 55
}