team-ownership
Assign and reassign CRM record ownership, audit who-owns-what across object types, and handle rep transitions. Built on `bulk-operations`.
Savant verdict: Review advised
Security analysis or source-hub review identified patterns requiring attention.
Live evaluation
Not evaluated yet. Workspaces can request a live evaluation.
Security assessment
Scan pending.
Structure
- No license declaredConfirm you may reuse this skill before importing it into your repository.
- The source hub's security checks flagged this skillSnyk: warn (Risk: MEDIUM · 1 issue)
SKILL.md
---
name: team-ownership
description: Assign and reassign CRM record ownership, audit who-owns-what across object types, and handle rep transitions. Built on `bulk-operations`.
triggers:
- "assign owner"
- "reassign records"
- "ownership audit"
- "rep leaving"
- "transfer records"
- "change owner"
- "find records owned by"
- "redistribute accounts"
---
Prereq: read `bulk-operations/SKILL.md` first. JSONL piping, pagination, dry-run/digest/confirm, and `hubspot history` recovery live there. Reshape patterns live in `bulk-operations/resources/json-patterns.md`.
`hubspot_owner_id` is a string field on `contacts`, `companies`, `deals`, and `tickets`. Owners are CRM users — `hubspot owners list` returns them; there is no `teams` object, so team-level views are client-side groupings by `hubspot_owner_id`.
## 1. Resolve email → owner ID
Never hardcode IDs — they are portal-specific. Resolve, then cache:
```bash
FROM_ID=$(hubspot owners list | jq -r 'select(.email=="sarah@company.com") | .id')
TO_ID=$(hubspot owners list | jq -r 'select(.email=="mike@company.com") | .id')
```
## 2. Find records for an owner
Same filter across all four object types. Add object-specific `--properties` for context. Unowned records use the `!property` form.
```bash
hubspot objects search --type contacts --filter "hubspot_owner_id=$FROM_ID" --properties email,firstname,lifecyclestage
hubspot objects search --type companies --filter "hubspot_owner_id=$FROM_ID" --properties name,domain
hubspot objects search --type deals --filter "hubspot_owner_id=$FROM_ID" --properties dealname,dealstage,amount
hubspot objects search --type tickets --filter "hubspot_owner_id=$FROM_ID" --properties subject,hs_pipeline_stage
# Records with no owner at all
hubspot objects search --type deals --filter "!hubspot_owner_id" --properties dealname,amount
```
>100 hits — page with the `--after` loop from `bulk-operations`. Counting only: pipe to `wc -l`.
## 3. Bulk reassign — search → update
Reshape each search row into `{id, properties:{hubspot_owner_id}}` and pipe to `objects update`. `objects update` is irreversible, so dry-run first: the dry-run emits a digest + `apply_command_hint` at every size (confirm = the record ID for one record, the row count for a batch). Then re-run with `--digest`/`--confirm` (see `bulk-operations/SKILL.md` § "Safe destructive workflow").
```bash
# Dry-run
hubspot objects search --type contacts --filter "hubspot_owner_id=$FROM_ID" \
| jq -c --arg to "$TO_ID" '{id, properties:{hubspot_owner_id:$to}}' \
| hubspot objects update --type contacts --dry-run \
| tee /tmp/reassign.preview.jsonl
# Execute — lift the digest + confirm from the preview line (required at every size)
digest=$(jq -r 'select(.digest != null) | .digest' /tmp/reassign.preview.jsonl)
confirm=$(jq -r 'select(.digest != null) | .target.id' /tmp/reassign.preview.jsonl) # single: record ID; batch: row count
hubspot objects search --type contacts --filter "hubspot_owner_id=$FROM_ID" \
| jq -c --arg to "$TO_ID" '{id, properties:{hubspot_owner_id:$to}}' \
| hubspot objects update --type contacts --digest "$digest" --confirm "$confirm"
```
Single-record assignment — no stdin, no jq. Still irreversible, so dry-run first (confirm = the record ID):
```bash
hubspot objects update --type contacts 12345 --property hubspot_owner_id=$TO_ID --dry-run
hubspot objects update --type contacts 12345 --property hubspot_owner_id=$TO_ID --digest <hash> --confirm 12345
```
## 4. Rep-leaves workflow
Loop over the four object types the rep touches:
```bash
FROM_ID=$(hubspot owners list | jq -r 'select(.email=="leaving@company.com") | .id')
TO_ID=$(hubspot owners list | jq -r 'select(.email=="taking-over@company.com") | .id')
for type in contacts companies deals tickets; do
echo "── $type ──"
hubspot objects search --type "$type" --filter "hubspot_owner_id=$FROM_ID" \
| jq -c --arg to "$TO_ID" '{id, properties:{hubspot_owner_id:$to}}' \
| hubspot objects update --type "$type" --dry-run
done
```
Review each digest line, then re-run per type with the `--digest`/`--confirm` lifted from that type's preview line (required at every size; confirm = the row count for a batch, the record ID for a single). Mis-reassigned? `hubspot history --since 1h` lists the affected IDs.
## 5. Team-level views (client-side grouping)
Group records by `hubspot_owner_id`, join to `owners list` for human-readable emails:
```bash
hubspot objects search --type deals --filter "dealstage!=closedwon AND dealstage!=closedlost" \
--properties hubspot_owner_id --format json \
| jq '.data | group_by(.properties.hubspot_owner_id)
| map({owner_id: .[0].properties.hubspot_owner_id, count: length})' \
> /tmp/by-owner.json
hubspot owners list \
| jq --slurpfile by /tmp/by-owner.json -r \
'. as $o | $by[0][] | select(.owner_id==$o.id) | "\($o.email)\t\(.count)"'
```