Month-close reconciliation prep
Categorize transactions against history and hand a human a clean exception list — it never files anything.
/loop cadence: monthly. Using the QuickBooks API (read), categorize the period's transactions against historical patterns. Append matches + anomalies to state-file month-close.md. Each round, build/refresh the exception list (uncategorized, unusual, likely-miscoded) for a human to review. Never post, file, or reconcile anything in QuickBooks. Stop when the exception list is complete; log errors and stop. Budget: cap $[X]/run. Hard cap: stop after 1 iteration per run.
claude-code
Implementation note
ADAPTED, not verbatim. Yellow loop: read-only prep + exception list; a human files everything.
More ops loops
SOP-drift catcher
Each week, compare how work actually happened against the written SOP and flag where they've drifted apart.
/loop cadence: weekly. Using the Notion MCP (read), compare recent task/project records against the documented SOP pages [LINKS]. Append drift findings (step, written vs actual, evidence) to state-file sop-drift.md. Each round, draft ONE SOP update proposal for the biggest drift — propose only, edit no live SOP. Stop after one proposal; log errors and stop. Budget: cap $[X]/run. Hard cap: stop after 1 iteration per run.
opslow risk
Unpaid-invoice chaser
Keep an aging ledger of unpaid invoices and draft the next nudge at each threshold — you send it.
/loop cadence: daily or weekly. Using the Stripe API (read), maintain an aging ledger of unpaid/overdue invoices. Append (customer, invoice, days overdue, threshold hit) to state-file invoice-aging.md. When an invoice crosses a threshold [7/14/30d], draft the appropriate reminder in the right tone — DRAFT ONLY. Never send outbound, never touch charges or refunds. Stop after drafting due nudges; log errors and stop. Budget: cap $[X]/run. Hard cap: stop after 1 iteration per run.
opshigh risk
KPI anomaly watch
Stay silent on normal days and page you with a pre-investigated cause only when a metric breaks its expected band.
/loop cadence: daily. Using the PostHog MCP and Stripe (read), check my key metrics [LIST] against their normal bands. Append daily readings to state-file kpi-watch.md. On a normal day, do nothing but log. When a metric breaks its band, pre-investigate (segment, correlate, likely cause) and draft an alert with the diagnosis. Never change data or send customer-facing messages. Stop after the check; log errors and stop. Budget: cap $[X]/run. Hard cap: stop after 1 iteration per run.
opslow risk