Automate weekly checkups on your agents
Stand up a **weekly-checkup agent** and wire it to **Scheduled Tasks** ([Run an agent on a schedule](/tutorials/run-an-agent-on-a-schedule)). Each run emits the same **weekly checklist** for agents, context, credentials, workflows, and replacement candidates, so rot surfaces on the calendar instead of during an incident.
Plus: three Admin-Agent passes: diff **last week vs this week** pasted notes with max six **CHANGED / NEW_RISK / STILL_OPEN** bullets, draft short **owner Slack bullets** from open rows, and tighten the cron prompt when **UNKNOWN** spam means the template needs narrower asks.
| Audience | Admins · Everyone |
|---|---|
| Time | ~12 min |
| Prerequisites | You already know where agents and schedules live ([Create an agent from scratch](/tutorials/create-an-agent-from-scratch), [Give your agent its job description](/tutorials/give-your-agent-its-job-description)). Cron wiring: [Run an agent on a schedule](/tutorials/run-an-agent-on-a-schedule). Helpful: [Audit and clean up your agents](/tutorials/audit-and-clean-up-your-agents) for cleanup mindset, [View your audit logs](/tutorials/view-your-audit-logs) when a scheduled run fails quietly, [Build your agent directory](/tutorials/build-your-agent-directory) when **who owns what** matters. |
| You'll end up with | One **Weekly Checkup** agent, frozen markdown sections **Agents / Context / Credentials / Workflows / Replacements**, one **weekly scheduled task** whose prompt **you** wrote, optional Discord reminder ([Connect Discord to your agents](/tutorials/connect-discord-to-your-agents)), and failures visible in **Audit Logs** ([View your audit logs](/tutorials/view-your-audit-logs)). |
When a tutorial shows italic text in quotation marks, it usually mirrors a label or helper string inside Auxot. Product copy changes between releases — if something reads differently in your workspace, trust what you see on screen.
Callouts with a Worth knowing gold accent are meant as must-read context before you move on. Blockquotes that open with Tip are lighter, optional depth.
Why this matters
Weekly checkups are boring until they are expensive. Stale context ships wrong answers (Keep your context files honest and fresh). Surprise credential swaps break tools (Rotate credentials without surprising your agents). Stuck workflows train people to bypass Auxot (Run a workflow). Zombie agents confuse newcomers (Replace an agent currently in use).
This is not the same lesson as frozen behavioral probes (Run scheduled canary checks on production agents). Same clock, different job. Canaries ask did answers degrade? Weekly checkups ask is your agent set documented, attached, and safe to operate?
Auxot does not silently inventory your whole account while you sleep. You configure the schedule. You optionally paste a short snapshot each week. The agent formats discipline, not omniscience. Missed weeks show up as empty sections, which is honest signal.
The calendar remembers. You still supply facts.
Quick start
- Mint Weekly Checkup: instructions: five sections, max six bullets each, UNKNOWN without paste, forbid invented usage stats.
- Paste baseline notes once: agent names and owners from your directory (Link your chat apps to your Auxot account). Redact secrets. Save the thread as reference.
- Lock weekly cron (Run an agent on a schedule): weekday morning, timezone explicit; prompt instructs the agent to emit a checklist plus OPEN_ROWS table (columns item, owner guess, severity).
- Optional Discord ping (Connect Discord to your agents): success channel posts the summary, error channel catches cron failures. Skim Audit Logs (View your audit logs) until trust earns shortcuts.
- Ten-minute human pass: walk Settings while reading the checklist, paste deltas back, archive markdown weekly.
Done? Four consecutive weekly exports (PDF or markdown) you can compare. STILL_OPEN count trends down, or you can explain every row that stayed open.
The agent can do that?
1. Week-over-week diff on pasted notes
Chat → Admin Agent:
Last week notes: […]. This week notes: […]. **CHANGED / NEW_RISK / STILL_OPEN** bullets only, max six, no repeat of resolved trivia.
Why it’s non-obvious: Full reruns waste attention. Leads want delta narrative after you paste both weeks.
2. Owner Slack bullets
OPEN_ROWS table: […]. Draft at most five Slack bullets, @owner placeholders OK, calm tense, ends with **Checkup log link** placeholder, no shame wording.
Why it’s non-obvious: Checkup work dies in private notes. Short channel posts recruit fixes (Update your agents without breaking the team).
3. Tighten the cron prompt against UNKNOWN spam
Example bad output: […]. Rewrite the scheduled prompt, 25 lines max, explicit paste asks, forbid guessing metrics, keep the five section headers literal.
Why it’s non-obvious: Loose prompts hallucinate traffic. Tight templates protect trust (Run an agent on a schedule).
Go deeper
Monthly deep audit
Pair with Audit and clean up your agents monthly. Run checkups weekly, audit at quarterly depth.
Credential rhythm
Checkup rows mention expiring integrations. Rotation choreography stays Rotate credentials without surprising your agents. Inventory lives in Manage your Credentials.
Canary contrast
When behavior regression worries you, schedule probes: Run scheduled canary checks on production agents. These checkups do not replace PASS or FAIL checks.
Leadership snapshot
Roll monthly checkup deltas into Run a quarterly review of your agents as an appendix. Steady artifacts beat rescue stories.
Walkthrough
Step 1: Pick Wednesday 9am local
Midweek beats Monday chaos. Adjust for your org.
Step 2: Dry-run manual before cron
Paste the same prompt by hand twice. Confirm stable headers. Edit the instructions once.
Step 3: Force a failure drill
Pause a credential intentionally in sandbox. Confirm the Error channel fires. Read Jobs (View your audit logs).
Step 4: Close one OPEN row live
Fix a credential naming typo. Paste a resolution note. Verify next week’s diff shows RESOLVED.
Step 5: Replace noise agents
The checkup flags no traffic + duplicate scope. Run the replacement checklist (Replace an agent currently in use). Do not delete casually.
Next steps
- → Run an agent on a schedule for the cron spine. This lesson is mostly what to schedule.
- → Save rubric-scored golden tests for your agents. Pair checklist checkups with numeric regression suites; same clock, sharper receipts.
- → Run scheduled canary checks on production agents to pair checkup output (roster health) with canaries (answer-shape health).
- → Audit and clean up your agents for deeper consolidation. These checkups keep gaps visible between audits.
- → Keep your context files honest and fresh when a checkup surfaces stale context rows.
- → Rotate credentials without surprising your agents when a checkup surfaces secret entropy.
Reference
- Pages in Auxot: Chat, Settings → Agents → Scheduled Tasks, Settings → Credentials, Audit Logs, optional Discord channels.
- Manual: Scheduled Tasks, Discord Integration.
- See also: Save rubric-scored golden tests for your agents, Enforce a living style guide and terminology, View your audit logs, Run a workflow, Manage your Credentials, Build your agent directory, Replace an agent currently in use, Connect Discord to your agents.