Audit My Data

Owners import contacts, half-fill opportunities, and never look back — so the data quietly rots: duplicate contacts, deals with no owner, tasks with no due date. This is the hygiene check-up. Read-only — it surfaces a fix list and offers the safe fixes one at a time.

Step 1 — Run the audits

Two read-only tools cover the data-hygiene surface; run both:

python ~/.claude/bos-run.py tool find-gaps --json
python ~/.claude/bos-run.py tool audit-contacts --json

• find-gaps.py — opportunities with no contact / no value / no stage / no owner; tasks overdue or with no due date.
• audit-contacts.py — missing/bad emails, likely-duplicate contacts (same email, or same name+company), dormant (365d+ no activity, no open opp), orphan (linked to nothing). --dormant-days N to tune.

If a script errors (auth/network), say so briefly and fall back to MCP reads (list_contacts, list_opportunities, list_tasks) — but that's many calls; prefer the scripts.

(For pipeline performance — stuck deals, stage drop-offs, value by stage — that's /weekly-review, not this skill. Keep this one to data quality.)

Step 2 — Present the report, worst first

Lead with what's actively costing money or will embarrass them, then tidy-ups.

🧹 Data audit — 3 things to fix, 2 worth a look

❌ FIX (costing you)
  → 4 opportunities have no owner — no one's working them. → assign
  → 2 likely-duplicate contacts (same email): "j.smith@…" appears twice. → merge/dedupe
  → 7 opportunities have no value set — your pipeline total is understated. → set value

⚠️ WORTH A LOOK
  → 12 contacts have no email — can't be emailed or blasted. → enrich or archive
  → 18 dormant contacts (no activity 365d+, no open deal). → archive or re-engage

✅ Clean: stages all placed, no orphan opportunities.

Use the operator's own pipeline/stage/field names (pull from the data, don't invent). Don't dump every row — show the worst few per category and the count.

On dormant-contact findings, name reactivation as the fix: a lead unworked for ~3 months has gone cold (business-method.md §10.3, directional), and a worked 200 beats a hoarded 5,000 — the value is in re-engaging the recent ones, not in keeping the pile.

Step 3 — Offer the fixes (with approval, one at a time)

For the safe, mechanical fixes, offer to apply them — one at a time, with a yes:

• Assign an unowned opportunity → mcp__trustpager__update_opportunity(id, assigned_user_ids=[...])
• Set a missing value → update_opportunity(id, amount=...) (confirm the number with the operator)
• Archive a dormant/orphan contact → confirm by name first.

Never merge or delete without explicit, named confirmation. Deduping is destructive — show the two records and which survives before doing anything. For bulk cleanups, do a small batch first, show the result, then continue.

What to never do

• ❌ Don't present findings alphabetically or as one flat list — rank by cost.
• ❌ Don't auto-fix — offer, get a yes, one at a time.
• ❌ Don't merge/delete records without showing them and naming what survives.
• ❌ Don't treat "dormant" as "delete" — it's a re-engage-or-archive decision.

Output shape

Headline tally, then FIX (worst first), then WORTH-A-LOOK, then a clean tally — and the single highest-value fix to make first.