AMCS Plan Generator

Transforms a Song Design Spec (SDS) into a deterministic execution plan that guides all downstream workflow nodes.

When to Use

Invoke this skill as the first node in the AMCS workflow when processing a new SDS. The plan establishes section order, target metrics, and evaluation criteria that inform Style, Lyrics, and Producer nodes.

Input Contract

inputs:
  - name: sds
    type: amcs://schemas/sds-1.0.json
    required: true
    description: Complete Song Design Spec containing all entity specs and constraints
  - name: seed
    type: integer
    required: true
    description: Determinism seed for this run

Output Contract

outputs:
  - name: plan
    type: amcs://schemas/plan-1.0.json
    description: |
      Execution plan containing:
      - section_order: Array of section names in performance order
      - target_word_counts: Per-section word count targets
      - evaluation_targets: Rubric thresholds from blueprint
      - work_objectives: Ordered list of tasks for downstream nodes

Determinism Requirements

• Seed: Use run seed directly (no RNG calls in this node)
• Temperature: N/A (no LLM generation required)
• Top-p: N/A
• Retrieval: None
• Hashing: Hash the complete plan output for provenance tracking

Constraints & Policies

• Section order MUST include at least one "Chorus" section
• Total target word count MUST respect sds.constraints.max_lines converted to approximate words (avg 6 words/line)
• Section requirements from sds.lyrics.constraints.section_requirements MUST be preserved
• If hook strategy is "lyrical" or "chant", require at least 2 chorus sections
• Duration targets MUST sum to within ±30 seconds of sds.constraints.duration_sec
• All blueprint-specified evaluation thresholds MUST be included in evaluation_targets

Implementation Guidance

Step 1: Extract Section Structure

1. Read sds.lyrics.section_order to get the base section sequence
2. Validate that at least one "Chorus" exists; if not, return error
3. If sds.lyrics.hook_strategy is "lyrical" or "chant", verify ≥2 chorus sections
4. Copy section order to plan.section_order

Step 2: Calculate Target Metrics

1. For each section in section_order:

   • Check sds.lyrics.constraints.section_requirements[section] for min/max lines
   • Convert lines to approximate word counts (6 words/line avg)
   • Store in plan.target_word_counts[section]

2. Validate total:

   • Sum all section word counts
   • Verify ≤ sds.constraints.max_lines * 6 words
   • If exceeded, proportionally reduce section targets

3. Duration targets:

   • If sds.producer_notes.section_meta contains target_duration_sec, copy to plan
   • Verify sum matches sds.constraints.duration_sec (±30s tolerance)

Step 3: Define Evaluation Targets

1. Load blueprint for sds.style.genre_detail.primary

2. Extract rubric thresholds:

   • hook_density: minimum acceptable score
   • singability: minimum acceptable score
   • rhyme_tightness: minimum acceptable score
   • section_completeness: minimum acceptable score
   • profanity_score: maximum acceptable score (0 if explicit=false)
   • total: minimum composite score

3. Store in plan.evaluation_targets

Step 4: Create Work Objectives

Generate ordered list of objectives for downstream nodes:

{
  "work_objectives": [
    {
      "node": "STYLE",
      "objective": "Generate style spec with tempo [min-max], key [primary], mood [list], enforcing blueprint tempo ranges and tag conflict matrix",
      "dependencies": []
    },
    {
      "node": "LYRICS",
      "objective": "Produce lyrics for sections [list], enforcing rhyme scheme [scheme], meter [meter], hook strategy [strategy]",
      "dependencies": ["STYLE"]
    },
    {
      "node": "PRODUCER",
      "objective": "Create production notes with [N] hooks, structure [structure], section tags from plan",
      "dependencies": ["STYLE"]
    },
    {
      "node": "COMPOSE",
      "objective": "Merge artifacts into render-ready prompt respecting [engine] character limits",
      "dependencies": ["STYLE", "LYRICS", "PRODUCER"]
    }
  ]
}

Step 5: Validate and Return

1. Validate plan against amcs://schemas/plan-1.0.json
2. Compute SHA-256 hash of plan JSON
3. Return plan with hash in metadata

Examples

Example 1: Christmas Pop Song

Input:

{
  "sds": {
    "style": {"genre_detail": {"primary": "Christmas Pop"}},
    "lyrics": {
      "section_order": ["Intro", "Verse", "PreChorus", "Chorus", "Verse", "Chorus", "Bridge", "Chorus"],
      "hook_strategy": "chant",
      "constraints": {
        "max_lines": 120,
        "section_requirements": {
          "Chorus": {"min_lines": 6, "max_lines": 10}
        }
      }
    },
    "producer_notes": {
      "hooks": 2,
      "section_meta": {
        "Intro": {"target_duration_sec": 10},
        "Chorus": {"target_duration_sec": 25}
      }
    },
    "constraints": {"duration_sec": 180, "max_lines": 120}
  },
  "seed": 42
}

Output:

{
  "section_order": ["Intro", "Verse", "PreChorus", "Chorus", "Verse", "Chorus", "Bridge", "Chorus"],
  "target_word_counts": {
    "Intro": 30,
    "Verse": 96,
    "PreChorus": 60,
    "Chorus": 54,
    "Bridge": 72
  },
  "evaluation_targets": {
    "hook_density": 0.7,
    "singability": 0.8,
    "rhyme_tightness": 0.75,
    "section_completeness": 0.9,
    "profanity_score": 0.0,
    "total": 0.8
  },
  "work_objectives": [
    {"node": "STYLE", "objective": "Generate Christmas Pop style with anthemic energy, brass instrumentation", "dependencies": []},
    {"node": "LYRICS", "objective": "Produce chant-heavy lyrics with AABB rhyme scheme across 8 sections", "dependencies": ["STYLE"]},
    {"node": "PRODUCER", "objective": "Create production notes with 2 hooks, lush mix, section-specific tags", "dependencies": ["STYLE"]},
    {"node": "COMPOSE", "objective": "Merge into Suno-compatible prompt under 5000 chars", "dependencies": ["STYLE", "LYRICS", "PRODUCER"]}
  ],
  "_hash": "abc123..."
}

Common Pitfalls

1. Missing Chorus: Failing to validate chorus presence causes downstream validation failures
2. Word Count Overflow: Not checking total against max_lines leads to truncated lyrics
3. Duration Mismatch: Ignoring target_duration_sec sum validation causes rendering issues
4. Blueprint Mismatch: Using wrong blueprint for genre results in inappropriate thresholds
5. Determinism Loss: Introducing any randomness breaks reproducibility guarantee