do-plan-critiquelisted
Install: claude install-skill tomcounsell/ai
# Plan Critique (War Room)
## Stage Marker
At the very start of this skill, write an in_progress marker:
```bash
sdlc-tool stage-marker --stage CRITIQUE --status in_progress --issue-number {issue_number} 2>/dev/null || true
```
After posting the verdict (Step 5), write the completion marker if READY TO BUILD, leave in_progress otherwise:
```bash
# On READY TO BUILD verdict:
sdlc-tool stage-marker --stage CRITIQUE --status completed --issue-number {issue_number} 2>/dev/null || true
```
## What this skill does
Critiques a plan document from six expert perspectives plus automated structural validation. Each critic has a defined lens and returns severity-rated findings. The skill aggregates, deduplicates, and produces a verdict: READY TO BUILD, NEEDS REVISION, or MAJOR REWORK.
## When to load sub-files
- Spawning war room critics → read [CRITICS.md](CRITICS.md) for critic definitions and prompt templates
## Quick start
1. Resolve the plan path from `$ARGUMENTS` (issue number or file path)
2. Read the plan and fetch linked issue/prior art context
3. Run automated structural checks (Step 2)
4. Spawn six parallel critics with the plan text (Step 3)
5. Aggregate findings and output the report (Steps 4-5)
## Plan Resolution
Resolve the plan document path from `$ARGUMENTS`:
```bash
ARG="$ARGUMENTS"
# If argument is a number, resolve from GitHub issue
if [[ "$ARG" =~ ^#?[0-9]+$ ]]; then
ISSUE_NUM="${ARG#\#}"
PLAN_PATH=$(gh issue view "$ISSUE_NUM" --json body -q '.bod