crlisted

# Dynamic Code Review Orchestrator > Relies on BDK foundation (STARTUP_INSTRUCTIONS.md) for project context and MCP tool preference. Code review orchestrator. Determine change scope, dispatch specialized agents parallel, collect results, merge into unified report. ## Terminal Output **On Start:** ``` ┌─────────────────────────────────────────────────┐ │ 👁️ ORCHESTRATOR: code-review │ │ 📋 Task: {brief description} │ │ ⚡ Model: sonnet │ └─────────────────────────────────────────────────┘ ``` **During Execution:** ``` [cr] Step 1: Determining scope... [cr] Scope: {N} files changed, {N} lines total → {tiny|small|large|massive} [cr] Step 2: Dispatching {N} agents... [cr] - {N}× layer-group reviewers (sonnet) [cr] - 1× architecture reviewer (opus) [cr] - 1× test reviewer (opus) [cr] - {N}× duplicate detectors (haiku) [cr] - 1× dead code detector (haiku) [cr] - 1× static analyse (haiku) [cr] - 1× test runner (haiku) [cr] Step 3: Waiting for agents... [cr] Step 4: Merging results... [cr] ✓ Complete ({N} findings: {critical}C/{high}H/{medium}M/{low}L) ``` ## Safety Rules (MANDATORY) - MUST NOT modify files. No Edit, Write, NotebookEdit tools. - All sub-agents read-only. ## Process ### Step 1: Determine Scope !`python3 ${CLAUDE_PLUGIN_ROOT}/scripts/inject.py --chain ${CLAUDE_PLUGIN_ROOT}/fragments/tool-tiers/review.chain.json` 1. Get risk-scored changed file list with severity ratings 2. I