memory-import-sanitizer

name	memory-import-sanitizer
description	Use this skill when importing ChatGPT or Claude exports, old Notion databases, local notes, or messy knowledge dumps into a durable knowledge system after redaction, deduplication, and classification. Do not use for live project planning, raw unredacted client data, or content that should remain out of Notion.

memory-import-sanitizer

Import AI conversation exports and external notes into your knowledge system after redaction, deduplication, and classification.

Purpose

When migrating from ChatGPT exports, old Claude conversations, Notion imports, or local markdown notes, clean and classify the content before it enters your system of record.

When to use this

• Importing ChatGPT conversation exports
• Migrating Claude conversation history
• Importing old Notion databases
• Consolidating local markdown notes into Notion
• Cleaning up a messy knowledge dump

When NOT to use this

• Do not use for live project planning, raw unredacted client data, or content that should remain out of Notion.

Process

1. Inventory the import

find /path/to/export -type f | head -50
wc -l /path/to/export/**/*.md

How many files? What format? How old? Any obvious junk?

2. Redaction pass

Remove before importing:

• API keys, tokens, passwords (even expired ones)
• Client names and confidential details (use client-boundary-guard rules)
• Personal information that shouldn't be in a shared system
• Credentials embedded in code snippets

3. Deduplication

• Same conversation exported multiple times? Keep the latest.
• Same decision recorded in 3 places? Keep the most detailed, note the others.
• Same code snippet in 5 conversations? Extract once, reference by location.

4. Classification

Sort into:

• Decisions → decision-log-adr format, store in Notion Decisions DB
• Runbooks/processes → format as runbook, store in Notion
• Code patterns → extract to skill or pattern library
• Project context → update relevant project's CLAUDE.md
• Stale/irrelevant → discard (most imports are 60%+ stale)

5. Import to Notion

For each classified item, create or update the appropriate Notion page. Link to source context where relevant.

Output format

# Import summary — [source]

## Inventory
- Total items: [N]
- After dedup: [N]
- After stale removal: [N]

## Classification
- Decisions: [N] → Notion Decisions DB
- Runbooks: [N] → Notion Runbooks
- Patterns: [N] → skill library / pattern-learner
- Project context: [N] → CLAUDE.md updates
- Discarded: [N] (stale/irrelevant)

## Redactions made
- [N] credentials removed
- [N] client references anonymized

## Follow-ups
- [ ] [Items needing human review before import]

Hard rules

• Never import without a redaction pass — assume exports contain secrets
• Never import duplicates — deduplicate first
• Most imports are majority stale — don't preserve everything out of completeness instinct
• Client data follows client-boundary-guard rules even in imports

Connects to

• client-boundary-guard — for redacting client-specific content
• notion-brain — destination for imported knowledge
• system-of-record-governance — imports must go to the right system
• decision-log-adr — for decisions found in imports

Common failure modes

Importing everything — 500 conversations, 450 are stale. Importing all of them pollutes your knowledge base. Be aggressive about discarding.

Credentials in old exports — ChatGPT exports from 2024 may contain API keys you pasted in. Always scan before importing.