astro44/bookend-score-complexity

bookend-score-complexity - Proposal Complexity Scoring

Evaluates proposal complexity against the scanned codebase to determine how many decomposition agents are needed and at what tier.

Input Schema

{
  "project_dir": "/path/to/project",
  "platform": "web|go|flutter|ios|python|rust",
  "proposal_text": "Full proposal or SPRINT_TODO.json content",
  "readiness_report": {},
  "bookends_dir": "/path/to/go-autonom8"
}

Instructions

1. Load Configuration

Read complexity thresholds from sprint_bookends.yaml:

• medium_at: 30
• complex_at: 55
• very_complex_at: 80
• max_agents: 5
• dimension_weights (if overridden)

2. Score Dimensions

Evaluate 6 complexity dimensions:

proposal_size (max 25): Word count of proposal text

• 2000+ words: 25, 900+: 18, 250+: 10, >0: 4

proposal_structure (max 8): Non-empty line count

• 80+ lines: 8, 30+ lines: 4

source_surface (max 25): Existing source file count

• 250+ files: 25, 80+: 18, 20+: 10, >0: 4

platform_surface (max 10): Platform manifest signal count

• 3+ signals: 10, 2 signals: 6

readiness_gap (max 15): Existing project with insufficient docs

• Existing + docs not OK: 15

requirement_domains (max 12): Domain keyword count in proposal

• 8+ domains: 12, 4+ domains: 7
• Keywords: accessibility, api, auth, browser, cache, database, deploy, design, mobile, migration, performance, security, terraform, test, visual, workflow

3. Classify Tier

Map total score to tier using sprint_bookends.yaml thresholds:

• score >= very_complex_at: very_complex
• score >= complex_at: complex
• score >= medium_at: normal
• else: simple

4. Determine Agent Load

• very_complex: 5 agents
• complex: 3 agents
• normal: 2 agents
• simple: 1 agent

Cap at max_agents from config.

Output Format

{
  "skill": "bookend-score-complexity",
  "status": "success",
  "score": 64,
  "tier": "complex",
  "suggested_agents": 3,
  "dimensions": [
    {"name": "proposal_size", "score": 25, "max": 25, "reason": "large proposal text"},
    {"name": "source_surface", "score": 18, "max": 25, "reason": "medium existing source surface"}
  ],
  "drivers": ["large proposal text", "medium existing source surface"]
}

Token Efficiency

• Pure text analysis + arithmetic, no LLM calls
• Sub-second execution
• Deterministic output for same inputs