# #1539 conversion: the agentic `flow-retrospective-cycle` skill expressed as a
# declarative YAML Flow. Like flow-architecture-evolution (#1547), the prose
# opens with a multi-agent fanout — Step 1 dispatches THREE pre-retro analysts
# in parallel (metrics, survey, patterns) then synthesizes them into the
# session agenda. That panel is encoded as a single `fanout` step. The
# SKILL.md remains the thin wrapper / discoverable trigger surface; this
# playbook is the orchestration source of truth for the step sequence + gate.
#
# Faithful to the prose flow's structure:
#   Step 1 pre-retro prep (×3 + agenda) → retro-prep        (FANOUT panel + agenda synthesis)
#   Step 2 facilitate session           → retro-facilitate  (scrum-master / agile-coach / analyzer)
#   (human gate: review session + prior actions) → action-review-gate
#   Step 3 create/prioritize actions     → retro-action-items (project-manager)
#   Step 4 document summary + tracker     → retro-document     (scrum-master)
#   Step 5 track effectiveness            → retro-track        (retrospective-analyzer)
apiVersion: flow.aiwg.io/v1
kind: FlowPlaybook
metadata:
  name: flow-retrospective-cycle
  labels:
    category: sdlc-orchestration
    domain: retrospective
spec:
  vars:
    retrospective_type: iteration
    iteration_number: ""
  steps:
    # Step 1: pre-retrospective preparation. Three analysts run in parallel —
    # metrics (scrum-master), survey design (agile-coach), pattern analysis
    # (retrospective-analyzer) — then the scrum-master synthesizes the agenda
    # from their outputs. Encoded as a fanout panel.
    - id: retro-prep
      fanout:
        strategy: parallel
        agents:
          - retro-prep-metrics    # scrum-master
          - retro-prep-survey     # agile-coach
          - retro-prep-patterns   # retrospective-analyzer
        synthesize: retro-prep-agenda   # scrum-master
      outputs:
        - name: agenda

    - id: retro-facilitate
      capability: retro-facilitate
      depends_on: [retro-prep]
      inputs:
        - { name: retrospective_type, from: "vars.retrospective_type" }
      outputs:
        - name: insights

    # Human checkpoint: facilitator + team review the session output and the
    # status of prior action items before committing to new ones. The retro is
    # only as good as the human judgement applied at this pause point.
    - id: action-review-gate
      kind: gate
      description: |
        Human gate (review): present the facilitated session insights, root
        causes, and the status of previous action items. The team confirms the
        improvement opportunities are real and prior actions are correctly
        marked (completed / in-progress / blocked) before new SMART action
        items are created. Per the prose Success Criteria: previous action
        items must be reviewed and updated here.
      depends_on: [retro-facilitate]

    - id: retro-action-items
      capability: retro-action-items
      depends_on: [action-review-gate]
      outputs:
        - name: action_items

    - id: retro-document
      capability: retro-document
      depends_on: [retro-action-items]
      outputs:
        - name: report_path

    - id: retro-track
      capability: retro-track
      depends_on: [retro-document]