Variant Pack Generation — Skill 82
Purpose
Generate multiple challenge instances simultaneously with controlled diversity. Enables weekly rotation, A/B testing, and contamination resistance.
Variant Pack Structure
{
"pack": {
"template_id": "tmpl-{family}-v{N}",
"family": "string",
"pack_size": 3-8,
"generated_at": "ISO-8601",
"cross_instance_similarity": "float (must be <0.70 for any pair)",
"instances": [
{
"instance_id": "BOUTS-2026-XXXX",
"seed": "8-char hex",
"mutations_applied": ["mutation_type"],
"framework": "string",
"database": "string",
"domain": "string",
"bug_types": ["string"],
"difficulty_profile": {},
"similarity_to_others": [0.0-1.0]
}
]
}
}
Pack Generation Rules
| Rule |
Requirement |
| Mutation diversity |
Each instance differs on at least 3 mutation types |
| Cross-instance similarity |
< 0.70 for any pair |
| Difficulty consistency |
Profiles within ±1 on each dimension across the pack |
| Shared attributes |
Same family, weight class, and format |
| Independent calibration |
Each instance gets its own calibration package |
| Similarity matrix |
Included in pack for publisher verification |
When to Generate Packs
| Context |
Pack Size |
Rationale |
| Weekly rotation |
4-pack |
1 per week, retire after use |
| Flagship families |
8-pack |
Long-term freshness for Blacksite/Fog of War/False Summit |
| Boss Fights |
1 (single) |
Monthly events are unique |
| Versus (asymmetric) |
2-pack |
One for each side |
| Seasonal events |
4-6 pack |
Cover the season with rotation |
| A/B testing |
2-pack |
Compare CDI between variants |
Quality Control
| Check |
Threshold |
Action if Failed |
| Any pair similarity > 0.70 |
Hard fail |
Regenerate the offending instance |
| Difficulty profile variance > 2 on any dimension |
Soft fail |
Constrain mutations, regenerate |
| Any instance fails calibration independently |
Hard fail |
Regenerate that instance only |
| Pack average CDI < B-Tier |
Soft fail |
Review template, consider redesign |
Pack Diversity Dimensions
Each instance in a pack should vary across:
| Dimension |
Example Variations |
| Framework |
Express → Fastify → Hono → Koa |
| Database |
PostgreSQL → MySQL → SQLite → MongoDB |
| Domain |
Payment processing → Inventory → Notifications → User management |
| Bug types |
Race condition → Deadlock → Connection pool → Memory leak |
| Red herring style |
Misleading TODO → Suspicious import → Dead code → Wrong error message |
| Log content |
Different timestamps, different error patterns, different noise level |
Integration Points
- Mutation Layer (Skill 52): Pack generation applies mutations per instance
- Calibration Packaging (Skill 81): Each instance gets its own calibration package
- Contamination Doctrine (Skill 49): Diversity is a contamination defense
- Challenge Economy (Skill 58): Packs feed weekly rotation and seasonal events
1---2name: variant-pack-generation3description: Variant Pack Generation — Skill 824---5# Variant Pack Generation — Skill 8267## Purpose8Generate multiple challenge instances simultaneously with controlled diversity. Enables weekly rotation, A/B testing, and contamination resistance.910## Variant Pack Structure1112```json13{14 "pack": {15 "template_id": "tmpl-{family}-v{N}",16 "family": "string",17 "pack_size": 3-8,18 "generated_at": "ISO-8601",19 "cross_instance_similarity": "float (must be <0.70 for any pair)",20 "instances": [21 {22 "instance_id": "BOUTS-2026-XXXX",23 "seed": "8-char hex",24 "mutations_applied": ["mutation_type"],25 "framework": "string",26 "database": "string",27 "domain": "string",28 "bug_types": ["string"],29 "difficulty_profile": {},30 "similarity_to_others": [0.0-1.0]31 }32 ]33 }34}35```3637## Pack Generation Rules3839| Rule | Requirement |40|------|-------------|41| Mutation diversity | Each instance differs on at least 3 mutation types |42| Cross-instance similarity | < 0.70 for any pair |43| Difficulty consistency | Profiles within ±1 on each dimension across the pack |44| Shared attributes | Same family, weight class, and format |45| Independent calibration | Each instance gets its own calibration package |46| Similarity matrix | Included in pack for publisher verification |4748## When to Generate Packs4950| Context | Pack Size | Rationale |51|---------|-----------|-----------|52| Weekly rotation | 4-pack | 1 per week, retire after use |53| Flagship families | 8-pack | Long-term freshness for Blacksite/Fog of War/False Summit |54| Boss Fights | 1 (single) | Monthly events are unique |55| Versus (asymmetric) | 2-pack | One for each side |56| Seasonal events | 4-6 pack | Cover the season with rotation |57| A/B testing | 2-pack | Compare CDI between variants |5859## Quality Control6061| Check | Threshold | Action if Failed |62|-------|-----------|-----------------|63| Any pair similarity > 0.70 | Hard fail | Regenerate the offending instance |64| Difficulty profile variance > 2 on any dimension | Soft fail | Constrain mutations, regenerate |65| Any instance fails calibration independently | Hard fail | Regenerate that instance only |66| Pack average CDI < B-Tier | Soft fail | Review template, consider redesign |6768## Pack Diversity Dimensions6970Each instance in a pack should vary across:7172| Dimension | Example Variations |73|-----------|--------------------|74| Framework | Express → Fastify → Hono → Koa |75| Database | PostgreSQL → MySQL → SQLite → MongoDB |76| Domain | Payment processing → Inventory → Notifications → User management |77| Bug types | Race condition → Deadlock → Connection pool → Memory leak |78| Red herring style | Misleading TODO → Suspicious import → Dead code → Wrong error message |79| Log content | Different timestamps, different error patterns, different noise level |8081## Integration Points8283- **Mutation Layer** (Skill 52): Pack generation applies mutations per instance84- **Calibration Packaging** (Skill 81): Each instance gets its own calibration package85- **Contamination Doctrine** (Skill 49): Diversity is a contamination defense86- **Challenge Economy** (Skill 58): Packs feed weekly rotation and seasonal events
nickgallick