| name | crm-migrate |
| description | One-shot migration between CRM backends. Markdown → Airtable, markdown → Attio, Airtable → markdown, Attio → markdown, or Airtable ↔ Attio. Use when first adopting an external CRM, when leaving one, or when swapping providers. Snapshot-style, not incremental. |
| disable-model-invocation | false |
| allowed-tools | Read, Write, Bash(mkdir *), Bash(cp *), Bash(date *) |
CRM Migrate
The migration tool. The adapter skills (crm-adapter-airtable, crm-adapter-attio) are for ongoing two-way sync — this skill is the one-time copy. Always snapshots the source and writes a migration log.
Inputs
$ARGUMENTS:
- Required:
--from=<markdown|airtable|attio> and --to=<markdown|airtable|attio>.
- Optional:
--dry-run.
- Optional:
--object=<companies|opportunities|outreach|all> — default all.
Examples
/career:crm-migrate --from=markdown --to=airtable
/career:crm-migrate --from=attio --to=markdown # leaving Attio
/career:crm-migrate --from=airtable --to=attio # swap providers
Procedure
1. Pre-migration snapshot
Always copy the source state first.
- Source = markdown: tar
crm/ → ${WORKING_FOLDER}/cache/crm-snapshot-<date>.tar.
- Source = airtable / attio: dump rows to
${WORKING_FOLDER}/cache/<source>-snapshot-<date>.json.
2. Read source
- markdown: parse tables to row arrays.
- airtable / attio: pull all records via the relevant adapter's read path.
3. Map to destination shape
Use the field mappings from the relevant adapter SKILL.md. For unmappable fields (e.g. moving from Attio's free-form Notes to markdown row notes column), concatenate with \n---\n and put in a single column.
4. Confirm plan
Show:
- N companies, N opportunities, N outreach rows to migrate.
- Any fields that will be lossy (e.g. linked-record collapses, custom attrs not in destination).
- Destination already has K rows — will the migration append (skip slug collisions) or overwrite? Ask.
--dry-run stops here.
5. Apply
Write to destination. For external CRMs, prefer the adapter's create-by-slug path so re-running is idempotent.
For markdown destinations: write tables in the canonical layout. Preserve any existing rows the migration didn't touch (don't clobber a markdown CRM the user was already curating).
6. Migration log
Write ${WORKING_FOLDER}/crm/.migration-<date>.md:
# CRM migration — {{date}}
- from: {{source}}
- to: {{dest}}
- objects: {{which}}
- snapshot: cache/<...>
- migrated: {{counts}}
- skipped: {{counts + reasons}}
- lossy fields: {{list}}
7. Print summary
migrated: companies=<n> opportunities=<n> outreach=<n>
skipped: <n> (see log)
snapshot: cache/<source>-snapshot-<date>.<ext>
log: crm/.migration-<date>.md
Guardrails
- Always snapshot before writing. Never destructive without a snapshot.
- Migrations are not incremental. After migration, switch to the relevant adapter for ongoing sync.
- If destination is non-empty, default to append (skip slug collisions) and prompt before overwrite.
Failure modes
- Mid-migration error → snapshot exists, partial writes are tolerated (idempotent on re-run by slug).
- Source unreadable → bail before writing destination.
