Card-News Orchestrator

Diagnose your card-news production state, select the right pipeline mode, and guide execution. No external tools — pure routing. Reads memory to recommend what to run and when.

---

Purpose

Card-News Orchestrator is the traffic controller for the 6 card-news skills. It answers:

• What card-news assets exist, and what's missing?
• Which mode fits my current request?
• Which skill should run next?
• What's the full execution path?

This skill never creates content itself. It reads memory to diagnose the current state, analyzes the user's request, and produces a step-by-step execution plan pointing to the right skills.

---

Memory Auto-Load Protocol

On every invocation, BEFORE any routing:

1. Check card-news-memory/ exists
   → If missing: create from card-news-memory-template/
   → If creation fails: proceed with all values as "NOT SET"
2. Try to read card-news-memory/ files:
   - series-config.md → Brand header, handle, tags, colors
   - production-log.md → Recent production history
   - copy-bank.md → Approved copy archive
   → Any file missing: mark as "NOT SET" in dashboard, continue
3. Optional: Check brand-memory/voice-profile.md → Brand tone
4. Optional: Check creative-memory/visual-guidelines.md → Visual identity
   → These directories may not exist — mark as "MISSING" and continue

어떤 파일이 없어도 orchestrator는 중단하지 않는다. Dashboard에 현재 상태를 표시하고, 사용자에게 누락된 설정을 안내한 뒤 가능한 모드로 진행한다.

---

Modes

Mode	Name	When to Use	Behavior
A	Full Production	Start from scratch — topic given, no existing copy	copy-writer → copy-evaluator → contents-manager → image-generator → card-news-maker
B	Re-render	Approved copy exists, need visual output	contents-manager → image-generator → card-news-maker
C	Copy Only	Write + evaluate copy, no rendering	copy-writer → copy-evaluator
D	Quick Render	All assets ready (copy + icons + images)	card-news-maker only

---

Process

Step 1: Build Production Status Dashboard

Generate from Auto-Load data:

Card-News Production Status
══════════════════════════════
Config:
  Series Config     : [CONFIGURED / NOT SET]
  Brand Header      : [value or MISSING]
  Account Handle    : [value or MISSING]

Recent Production:
  Last produced     : [DATE or NEVER]
  Total sets        : [N]

Assets:
  Copy Bank entries : [N] (few-shot references)

External Context:
  Brand Voice       : [AVAILABLE / MISSING]
  Visual Guidelines : [AVAILABLE / MISSING]
  Research Data     : [AVAILABLE / MISSING]

Always show the dashboard first — it grounds the conversation in facts.

---

Step 2: Gather Input

Collect from the user:

Input	Required	Default	Notes
Topic / Theme	Yes	—	The subject of the card-news
Card count	No	4	Range: 2~10
Language	No	한국어	Copy language
Card structure	No	Auto (copy-writer decides)	Override: explicit type sequence
Color override	No	series-config defaults	Custom --highlight, --feature-bg, etc.

If series-config.md has no brand header or handle configured, prompt the user to set them before proceeding. Write to series-config.md if new values are provided.

---

Step 3: Select Mode

Condition	Auto-Suggest
New topic, no existing copy	Mode A: Full Production
Copy exists (from copy-bank or user-provided), needs rendering	Mode B: Re-render
User wants copy only ("카피만", "글만 써줘")	Mode C: Copy Only
All assets provided (copy + images + icons)	Mode D: Quick Render
Ambiguous	Show dashboard → ask user to pick

Present the suggested mode with a one-line rationale. Let the user confirm or override.

---

Step 4: Build Execution Path

Skill Dependency Map

              ┌──────────────┐
              │ orchestrator │  ← you are here
              └──────┬───────┘
                     │
        ┌────────────┼────────────┐
        │            │            │
        ▼            ▼            │
 ┌────────────┐ ┌──────────────┐ │
 │ copy-writer│ │contents-mgr  │ │
 └─────┬──────┘ └──────┬───────┘ │
       │               │         │
       ▼               │         │
┌───────────────┐      │         │
│copy-evaluator │      │         │
└───────┬───────┘      │         │
        │ PASS         ▼         ▼
        │     ┌──────────────┐ ┌──────────────┐
        └────>│card-news-maker│ │image-generator│
              └──────────────┘ └──────────────┘

Mode A: Full Production

Full Production Pipeline
══════════════════════════
Step 1: copy-writer        → Card structure + all copy
Step 2: copy-evaluator     → Quality gate (PASS / REVISION)
        ↳ REVISION → back to copy-writer with feedback
Step 3: contents-manager   → Icon selection + image prompts
Step 4: image-generator    → AI image generation (nanobanana)
Step 5: card-news-maker    → HTML assembly + Playwright render → PNG

Note: Steps 1-2 and Step 3 can run in parallel.
      Step 4 needs Step 3 output. Step 5 needs all previous outputs.

Mode B: Re-render

Re-render Pipeline
══════════════════════════
Step 1: contents-manager   → Icon selection + image prompts
Step 2: image-generator    → AI image generation
Step 3: card-news-maker    → HTML assembly + render → PNG

Mode C: Copy Only

Copy Only Pipeline
══════════════════════════
Step 1: copy-writer        → Card structure + all copy
Step 2: copy-evaluator     → Quality gate
        ↳ REVISION → back to copy-writer

Mode D: Quick Render

Quick Render Pipeline
══════════════════════════
Step 1: card-news-maker    → HTML assembly + render → PNG

---

Step 5: Guide Execution

Initial handoff:

▶ Next skill: [SKILL_NAME]
  Purpose: [one-line description]
  Input needed: [what to prepare]

Say "[SKILL_NAME] 실행해줘" to start.

After each skill completes:

✅ [COMPLETED_SKILL] done → [output description]

▶ Next: [NEXT_SKILL_NAME]
  Purpose: [one-line description]

Continue? Say "[NEXT_SKILL_NAME] 실행해줘"

After all skills complete:

Summarize what was produced:

• Number of cards rendered
• Output path
• Log entry in production-log.md

Suggest next options: make another set, modify copy, change theme, export for different platform.

---

Skill Quick-Reference

Skill	MCP Tool	Output
copy-writer	None	Structured copy markdown
copy-evaluator	None	PASS / PASS WITH NOTES / REVISION
contents-manager	None	Icon SVGs + image prompts
image-generator	nanobanana	PNG image files
card-news-maker	Playwright	Assembled HTML + card PNG files (1080x1350)

---

Cross-Plugin Connections

Direction	Connection	Purpose
← brand-memory	voice-profile.md → copy-writer, copy-evaluator	Brand voice consistency
← creative-memory	visual-guidelines.md → contents-manager, image-generator	Visual identity
← research-memory	topic research → copy-writer	Research-backed content
→ creative-mkt	card images → image-creator for variations	Visual asset reuse

---

What This Skill Does NOT Do

• Write copy → copy-writer
• Evaluate copy → copy-evaluator
• Plan visual assets → contents-manager
• Generate images → image-generator
• Render HTML/PNG → card-news-maker
• Create content of any kind → It only routes and coordinates

Card-News Orchestrator stays focused on routing — diagnosing, planning, and guiding.