Badcase Archive

This directory contains low-quality intake issues that were rejected but preserved for training and improvement.

Purpose

1. Training data — Use rejected cases to improve the scoring algorithm
2. Pattern recognition — Identify common low-quality patterns
3. Contributor guidance — Generate improvement suggestions
4. Avoid repetition — Prevent same mistakes in future submissions

Directory Structure

badcase/
├── README.md                 # This file
├── index.json               # Index of all badcases
├── {issue-number}/          # One directory per issue
│   ├── intake.md           # Original intake content
│   ├── metadata.json       # Scoring details
│   └── reasons.md          # Rejection reasons

Categories

Issues are categorized by rejection reason:

Category	Description	Count
incomplete	Missing required sections	—
vague	Too short or unclear	—
test	Test/heartbeat submissions	—
spam	Promotional or off-topic	—

How It Works

Auto-Archive (score < 40)

When an intake issue receives a score below 40, it is automatically archived here with:

• intake.md — The original content
• metadata.json — Detailed scoring breakdown
• reasons.md — Specific rejection reasons and suggestions

Algorithm Improvement

Badcases are used to:

1. Extract features — What makes a low-quality submission
2. Adjust weights — Tune scoring dimensions
3. Add rules — Create new rejection patterns
4. Generate tests — Create test cases for validation

Pattern Analysis

Run analysis on badcases:

python3 scripts/analyze_badcases.py

This generates:

• Common rejection reasons
• Feature importance rankings
• Suggested weight adjustments

Improvement Suggestions

Each badcase includes specific suggestions:

## Suggestions for Improvement

- Add an 'Error' section with error messages
- Include code blocks with examples
- Add verification steps in '## Verification'
- Remove user-specific paths
- Add more technical detail

Permanent Retention

Badcases are permanently retained as:

• Training data for algorithm improvement
• Reference examples for contributor guidance
• Historical record of submission quality

Maintenance

• Monthly: Review new badcases for patterns
• Quarterly: Update improvement suggestions
• Annually: Archive old badcases to cold storage