Memory Persistence

When to Use

Activated whenever pod-research emits a GO decision or when users explicitly request saving a design/pattern. Ensures deterministic storage for later recall.

Execution Flow

1. On pod-research GO:
   • Append the full result to .claude/memories/validated_niches.json.
   • Respect 90-day archive rule and cap per category (top 100 by confidence).
2. On pod-research SKIP:
   • Archive to .claude/data/skipped_niches.json for pattern spotting (no auto-influence yet).
3. Provide query docs for standard shells (jq, grep) referencing stored files.
4. Trigger cleanup: remove entries older than 90 days and keep highest confidence 100 per brand.

Output Format

{
  "niches": [
    {
      "niche": "indoor plant care",
      "decision": "GO",
      "confidence": 0.85,
      "etsy_count": 18500,
      "trend_score": 62,
      "brand": "LWF",
      "date_validated": "2025-12-07",
      "notes": null
    }
  ]
}

Performance Target

• Token budget: 500–900 per memory write.
• Deterministic accuracy: 100% writes when GO, archive SKIP.
• Cleanup: Auto-archive entries older than 90 days, limit 100 per brand by highest confidence.
• Query guidance: CLI examples included within SKILL.md for future use.