meridianreviewlisted

# /meridian:review — Two-Stage Code Review Run spec compliance + code quality review on completed work. ## Arguments - `--phase <id>` — Review specific phase (default: current phase) - `--stage <1|2>` — Run only one stage - `--files <paths>` — Review specific files instead of full phase - `--cross-model` — Run independent review from a secondary AI model after Stage 2 - `--persona <name>` — Apply role-typed review lens (pm, architect, ux, qa, security) - `--party` — Concurrent multi-perspective review: code quality, security, and UX in parallel ## Procedure ### Step 0A: Party Mode (if --party) If `--party` is passed, skip the normal Stage 1/2 pipeline and run this instead. #### 0A-1: Build reviewer prompts for all 3 perspectives ```bash PYTHONPATH=$MERIDIAN_HOME uv run --project $MERIDIAN_HOME -- python -c " import json from scripts.party_review import build_reviewer_prompt, list_perspectives perspectives = list_perspectives() print(json.dumps(perspectives, indent=2)) " ``` Get changed files (same as Step 2 below), then build one prompt per perspective: ```bash PYTHONPATH=$MERIDIAN_HOME uv run --project $MERIDIAN_HOME -- python -c " from scripts.party_review import build_reviewer_prompt prompt = build_reviewer_prompt( '<perspective_key>', changed_files=['<file1>', '<file2>'], phase_name='<name>', phase_description='<desc>', ) print(prompt) " ``` #### 0A-2: Launch 3 reviewers in parallel Spawn 3 independent Agents (subagent_type: general-purpose) simul