Memory Export

2. Build Export Package

• Create a single JSON export object with this structure:

  {
    "metadata": { "exported_at", "source_project", "entry_counts", "total_confidence", "version" },
    "patterns": [ ...all pattern entries... ],
    "preferences": [ ...all preference entries... ],
    "errors": [ ...all error entries... ],
    "context": [ ...all context entries... ],
    "archive": [ ...all archived entries... ]
  }
  

• Load every entry JSON file from each subdirectory
• Include the full entry object (not just index references)
• Set version to "1.0" for format compatibility tracking

3. Write Export File

• Default export path: memory/exports/memory_export_{YYYYMMDD_HHmmss}.json
• Create memory/exports/ directory if it does not exist
• Write the JSON with 2-space indentation for readability
• Calculate and append SHA256 checksum as metadata.checksum

4. Verify Export Integrity

• Read the exported file back
• Parse JSON and verify all sections are present
• Compare entry counts against the inventory from step 1
• Confirm checksum matches

5. Import (Restore) Procedure

• When restoring: read the export JSON file
• For each entry in each section, check for existing entry by id
• If entry exists: skip (or overwrite if --force flag set)
• If entry is new: write to appropriate memory/{type}/{id}.json
• Rebuild all index.json files after import
• Log import action with counts: imported, skipped, overwritten

6. Report

• Output export/import summary: file path, entry counts per type, total size
• Log action to logs/claude_log.md with timestamp

Output

• memory/exports/memory_export_{timestamp}.json — full memory dump
• Integrity verification result (pass/fail)
• Export/import summary with per-type counts
• Log entry in logs/claude_log.md

Review Checklist

• All subdirectories included in export (patterns, preferences, errors, context, archive)
• Full entry objects exported, not just index references
• Metadata includes export date, counts, and checksum
• Export file verified by read-back and count comparison
• Import handles duplicates gracefully (skip or overwrite)
• Indexes rebuilt after import