/validate

Goal

Run /project-management:validate to ensure a project matches the standard blueprint and produce a scored report.

Preparation

1. Ask for root path (default current).
2. Determine whether to fix_issues (y/n) and if strict_mode should flag extra directories/files.
3. Load scoring weights from ../config/config.json -> validation.scoring_weights.

Execution Steps

1. Directory/File Scan

   root="<abs path>"
   required_dirs=(claude-code data/raw data/cleaned codes paper/figures paper/manuscripts pre/poster pre/slides)
   required_files=(README.md .gitignore project.yml)
   

   • Capture missing and extra items (ignore .git, .claude, .claude-plugin, __pycache__).
2. Score Calculation
   • dir_score = (required_dirs - missing_dirs)/required_dirs * 40
   • file_score = (required_files - missing_files)/required_files * 35
   • content_score = 15 if key files exist and non-empty else 0.
   • git_score = 10 when .git exists and has commits (git rev-list --count HEAD > 0).
   • Total score = round sum to int.
3. Issue List
   • Missing directories/files.
   • Strict mode: list extras.
   • Empty or tiny files (<10 bytes) flagged for rewrite.
4. Auto-Fix (optional) When fix_issues=true, create missing directories and render templates with missing_only=true so existing files remain untouched. Document each fix.
5. Report Respond with:
   • Score out of 100 + qualitative tier (>=90 excellent, >=75 good, >=60 fair, else poor).
   • Table/bullets of missing vs. extra.
   • Fixes applied or pending.
   • Recommendations (e.g., “Initialize Git before sharing”).

Error Handling

• Root missing → abort with helpful tip.
• If Git commands fail, treat as warning and continue scoring without Git points.
• When template rendering fails, include failing file path and leave instructions for manual creation.

Example Reply

/Users/adrian/projects/causal-impact scored 82/100 (good)
Missing: pre/poster, README.md
Extras (strict): old-data/
Fixes: created pre/poster, generated README.md from template.
Recommendation: move raw datasets under data/raw and remove old-data after backup.