Docs

Project documentation and wiki files

/cron-jobs.md

cron-jobs

cron/cron-jobs.md

Cron Job Registry

Purpose: Single source of truth for every active cron job in the crew. Source data: ~/.openclaw/cron/jobs.json (master) + ~/.openclaw/workspace/cron/jobs.json (workspace copy) Last updated: 2026-05-10 (t-1015: auto-dispatch → 5m, agent-poll-oscar added) Maintained by: Kelly 🌺 Related:

— failure detection, triage, post-mortems

— daily learning aggregation

Cron job state — runtime state including last run, next run, consecutive errors

— historical health tracking (deprecated)

Active Cron Jobs#

🌅 Morning Digest#

Field	Value
Job ID	`4695c6ea-ea22-4d33-9f15-f42f4acefc02`
Name	`morning-digest`
Agent	main (Robbie 🧐)
Schedule	Daily at 09:00 SGT (`0 9 * * *` Asia/Singapore)
Session target	Isolated
Delivery	WhatsApp → +6580860085
Enabled	✅ Yes
Success indicator	WhatsApp message delivered with task summary, blockers, and recent failures
Last run status	✅ OK (69s)
Description	Generates a crew digest (task summary, blockers, failed runs, upcoming jobs) and sends it to Maxime's WhatsApp. Max 20 lines, plain text, no markdown.

📖 Kelly Daily Docs Review#

Field	Value
Job ID	`acfc08aa-a69b-4a57-9304-007450ce1fb0`
Name	`kelly-daily-docs-review`
Agent	Kelly 🌺
Schedule	Daily at 03:00 SGT (`0 3 * * *` Asia/Singapore)
Session target	Isolated
Delivery	None (posts to Discord #kelly)
Enabled	✅ Yes
Success indicator	All watched repos reviewed; docs updated or confirmed current; `lastReviewedSha` updated; summary written to daily memory file
Last run status	✅ OK (473s)
Description	Kelly's primary job — reviews all watched repos for code changes, updates or creates docs, commits changes, and logs the summary. Timeout: 1200s.

📊 Kelly Weekly Index Sweep#

Field	Value
Job ID	`00f508b5-823c-4a34-ab7c-3b4a48bb534f`
Name	`kelly-weekly-index-sweep`
Agent	Kelly 🌺
Schedule	Sunday at 02:00 SGT (`0 2 * * 0` Asia/Singapore)
Session target	Isolated
Delivery	None (posts to Discord #kelly)
Enabled	✅ Yes
Success indicator	`docs/INDEX.md` verified, all projects checked, new projects added, baseline doc check complete
Last run status	⚪ Not yet run (created 2026-05-08, first Sunday is 2026-05-10)
Description	Weekly cross-project index maintenance: verify INDEX.md accuracy, add new projects, flag broken paths, check baseline docs exist. Timeout: 600s.

✅ Kelly Task Quality Gate (Every 15min)#

Field	Value
Job ID	`e2bcff6e-f35f-4dca-be0f-d302d45128c7`
Name	`kelly-task-qc-15min`
Agent	Kelly 🌺
Schedule	Every 15 minutes (`/15 * * *` Asia/Singapore)
Session target	Isolated (light context)
Delivery	WhatsApp → +6580860085 (only if tasks were QC'd)
Enabled	✅ Yes
Success indicator	Scans all inbox/needs_review tasks; QC found nothing OR tasks quality-gated to reviewState: kelly_done + status set to backlog + relayed to Robbie
Last run status	✅ OK (freshly registered 2026-05-09)
Description	Scans all task files for those with `status: inbox` (need QC) or `status: needs_review` (errant). Fixes: truncated titles, missing tags, missing priority, auto-triage suggested_assignee, priority mismatches (title says P0/P1 but priority field says P2). Sets `reviewState: kelly_done`, `reviewRequested: true`, fixes status to `backlog`. Relays to Robbie via WhatsApp if any were processed. Timeout: 300s, light context.

🗃️ Mission Control Git Sync#

Field	Value
Job ID	`6745d740-ec5d-4970-a15c-18b3c591e675`
Name	`mission-control-git-sync`
Agent	main (Robbie 🧐)
Schedule	Daily at 02:00 SGT (`0 2 * * *` Asia/Singapore)
Session target	Isolated
Delivery	WhatsApp → +6580860085 (announce)
Enabled	✅ Yes
Success indicator	Git auto-commit + push completes; or clean exit with "No changes to commit"
Last run status	✅ OK (15s)
Description	Staged-file git sync for `/home/maxime/projects/mission-control`. Runs `git add -A`, commits with changelog, and pushes to `Max-src/mission-control.git main`.

💾 Workspace Daily Backup Sync#

Field	Value
Job ID	`46198197-c281-490d-8c81-4d0114b84f9d`
Name	`workspace-daily-sync`
Agent	main (Robbie 🧐)
Schedule	Daily at 02:00 SGT (`0 2 * * *` Asia/Singapore)
Session target	Isolated
Delivery	None (failure notification only via WhatsApp)
Enabled	✅ Yes
Success indicator	Backup script runs cleanly; `~/.openclaw-backup/.last-sync.json` has `status=ok`
Last run status	✅ OK (16s)
Description	Full workspace backup via `scripts/backup-sync.sh`: rsync mirroring, config sanitization, secret scan, git commit+push to `Max-src/openclaw-workspace`. Timeout: 300s.

🔍 Vera Code Review (Every 2h)#

Field	Value
Job ID	`08d840e5-0938-485a-ac74-46c81fd10ac1`
Name	`vera-code-review-2h`
Agent	Vera 🔍
Schedule	Every 2 hours (`0 /2 * *` Asia/Singapore, staggerMs: 900000)
Session target	Isolated
Delivery	Discord → #chris channel (`1502183280676438016`)
Enabled	✅ Yes
Success indicator	New commits reviewed and posted to Discord; or silent exit if no new commits
Last run status	✅ OK (12s)
Description	Checks mission-control main for new commits since last review. If there are changes, does a full code review and posts findings to Discord #chris. Staggered 15min from the hour to avoid contention. Timeout: 300s, light context.

🔍 Vera Weekly Quality Deep Dive#

Field	Value
Job ID	`4e0cc4ea-b5a7-4997-bb93-6d9fae2384d1`
Name	`vera-weekly-quality-deep-dive`
Agent	Vera 🔍
Schedule	Sunday at 10:00 SGT (`0 10 * * 0` Asia/Singapore, staggerMs: 1800000)
Session target	Isolated
Delivery	Discord → #chris channel (`1502183280676438016`)
Enabled	✅ Yes
Success indicator	Weekly quality report posted to Discord #chris with recurring issues surfaced
Last run status	⚪ Not yet run
Description	Scans task completions, run logs, and mission-control commits for the past week. Surfaces recurring code quality patterns. Staggered 30min from the hour. Timeout: 600s, light context.

🔍 Vera Task Pipeline Audit (Daily at 03:00)#

Field	Value
Job ID	`3c8e9a65-9813-4ec2-a533-07cc6980bcd6`
Name	`task-pipeline-audit`
Agent	Vera 🔍
Schedule	Daily at 03:00 SGT (`0 3 * * *` Asia/Singapore)
Session target	Isolated (light context)
Delivery	Announce (last channel)
Enabled	✅ Yes
Success indicator	Script runs cleanly; legacy statuses normalized; warnings flagged; errors escalated to Robbie
Last run status	⚪ Not yet run (created 2026-05-09)
Description	Runs `scripts/task-pipeline-audit.py --fix` daily at 03:00 SGT. Scans all task files for skipped pipeline stages (`done` without prior `in_progress`), unknown/invalid statuses (`needs_review`, `running`, etc.), and auto-normalizes legacy statuses (`success`→`done`, `failed`→`in_progress`, `completed`→`done`). If errors found (tasks stuck with truly invalid states), escalates to Robbie via WhatsApp. Timeout: 120s, light context.

🔥 Oscar Daily Cost Check#

Field	Value
Job ID	`1cc33ed4-d499-43f2-94c0-1551e3693507`
Name	`oscar-daily-cost-check`
Agent	Oscar 🔥
Schedule	Daily at 08:00 SGT (`0 8 * * *` Asia/Singapore)
Session target	Isolated
Delivery	WhatsApp → +6580860085 (best effort)
Enabled	✅ Yes
Success indicator	Cost check script output delivered to Maxime's WhatsApp
Last run status	✅ OK (13s)
Description	Runs `python3 /home/maxime/.openclaw/workspaces/oscar/cost-check.py` and delivers the stdout to Maxime. Light context. Timeout: 120s.

🦑 Daily Reflection Aggregation#

Field	Value
Job ID	`b0ba16bc-7fac-4025-864d-b443cc024bca`
Name	`daily-reflection-aggregate`
Agent	Debbie 🦑
Model	`deepseek/deepseek-chat` (DeepSeek Flash)
Schedule	Daily at 23:30 SGT (`30 23 * * *` Asia/Singapore)
Session target	Isolated
Delivery	WhatsApp → +6580860085 (best effort)
Enabled	✅ Yes
Success indicator	All agents' reflections read; daily summary written to `team-management/reflections/daily/`; learning.md and patterns.md updated; WhatsApp summary sent
Last run status	❌ Failed — timeout (120s). Agent instruction set is complex (600s timeout needed).
Description	Reads each agent's daily reflection file, produces a consolidated crew learning summary, updates patterns.md and learning.md, and relays a WhatsApp summary. Uses DeepSeek Flash model (t-313: rerouted from Qwen — analysis-heavy task requires coherent multi-file summaries). Timeout: 600s, light context. Note: Last run timed out at 120s — timeout may need adjustment.

🦑 Fallback Monitor (Hourly)#

Field	Value
Job ID	`814c3663-750b-4a36-8130-e05fd90ebdf4` (master) / `f1a0e1c4-f97b-4fea-8b54-f6572be0c37c` (workspace copy)
Name	`fallback-monitor-hourly`
Agent	Debbie 🦑
Schedule	Every hour (`0 * * * *` Asia/Singapore, staggerMs: 300000)
Session target	Isolated
Delivery	None (sends WhatsApp alerts for new fallback events)
Enabled	✅ Yes
Success indicator	Script runs; new fallback events are alerted via WhatsApp; summary appended to `/tmp/mc-daily-summary.log`
Last run status	⚪ Not yet run (added 2026-05-09, first run pending)
Description	Runs `scripts/fallback-monitor.py` to scan task run logs for API rate-limit/fallback events. Reads `state/fallback-state.json`, delivers pending alerts via WhatsApp, marks them delivered. Staggered 5min from the hour. Timeout: 120s.

🦑 Rate-Limit Recovery Probe (Every 15min)#

Field	Value
Job ID	`0c99eea3-5b44-41c0-9daa-51e4f14efa83` (master) / `d6820459-1f31-40f9-a930-69891342a65f` (workspace copy)
Name	`rate-limit-recovery-probe`
Agent	Debbie 🦑
Schedule	Every 15 minutes (`/15 * * *` Asia/Singapore)
Session target	Isolated
Delivery	None (sends WhatsApp alerts on recovery detection)
Enabled	✅ Yes
Success indicator	Script runs; recovered models detected and alerted; summary appended to `/tmp/mc-daily-summary.log`
Last run status	✅ OK (25s)
Description	Runs `scripts/rate-limit-recovery.py` to probe rate-limited models. When a model recovers, sends WhatsApp alert and clears recovery status in `fallback-state.json`.

🦑 Review Flag Scan (Hourly)#

Field	Value
Job ID	`a6023281-6620-44e4-a396-72862868dcae`
Name	`review-flag-scan-1h`
Agent	Debbie 🦑
Schedule	Every hour (`0 * * * *` Asia/Singapore, staggerMs: 300000)
Session target	Isolated
Delivery	WhatsApp → +6580860085 (best effort, only if review-flagged tasks found)
Enabled	✅ Yes
Success indicator	Scans all tasks for `reviewRequested=true`; reports to WhatsApp if any found; silent exit (NO_REPLY) if none
Last run status	✅ OK (16s)
Description	Checks every task file for `reviewRequested=true` in frontmatter. If any found, delivers report via WhatsApp. Staggered 5min from the hour to avoid collision with fallback monitor.

🦑 Adoption Verifier (Hourly)#

Field	Value
Job ID	`c02a97f3-b4a1-4464-be1c-6b7e2c25b412`
Name	`adoption-verifier-hourly`
Agent	Debbie 🦑
Schedule	Every hour (`0 * * * *` Asia/Singapore, staggerMs: 300000)
Session target	Isolated
Delivery	WhatsApp → +6580860085 (announce, best effort; only sends if tasks reopened)
Enabled	✅ Yes
Success indicator	Pending adoption checks scanned; tasks with real output detected stay done; tasks without output reopened as blocked and alerted to WhatsApp
Last run status	⚪ Not yet run (added 2026-05-09, first run pending)
Description	Runs `scripts/adoption-verifier.py run-pending` to scan all pending adoption checks (process-design tasks marked done ~24h prior). For each: detects real process output (related tasks, files in reflections/state/reports/, enabled cron jobs). If output found → verified. If none → reopens task as `blocked` and alerts WhatsApp. Staggered 5min from the hour. Timeout: 60s.

Related:

Script: scripts/adoption-verifier.py
Hook: scripts/adoption-on-complete.sh (called during task lifecycle stage 6 to schedule checks)
State: state/adoption-checks.json
Lifecycle: PLAYBOOK.md §9.1 (Adoption Verification), §2 (Task Lifecycle stage 6.6)
Pattern: reflections/patterns.md (process design ≠ process adoption)

🦑 Team Health Check (Every 15min)#

Field	Value
Job ID	`9e548caf-9a7c-4815-9743-2eb7308b4b98`
Name	`team-health-every-15min`
Agent	Debbie 🦑
Model	`deepseek/deepseek-chat` (DeepSeek Flash)
Schedule	Every 15 minutes (`/15 * * *` Asia/Singapore)
Session target	Isolated
Delivery	WhatsApp → +6580860085 (announce, best effort; only sends if stale_alert_hours threshold exceeded or failures detected)
Enabled	✅ Yes
Success indicator	Health checks run every 15min independently of any session. Stale tasks, backlog rot, blocked tasks, unassigned tasks, and failures detected. WhatsApp alerts sent when stale_alert_hours (default 48h) exceeded. Quiet hours (23:00-08:00 SGT) respected. State logged to Debbie's daily memory and `state/team-health-state.json`.
Last run status	⚪ Not yet run (created 2026-05-09, first run pending)
Description	Runs the team health check protocol: fetch all tasks via API, scan for stale tasks (>24h todo, >72h backlog), blocked dependencies, unassigned tasks, and recent failures. Alerts WhatsApp when stale_alert_hours (48h) threshold is breached. Uses DeepSeek Flash model (t-313: rerouted from Qwen — analysis-heavy task needs root cause understanding). Runs every 15min with 120s timeout, isolated session. Previously embedded in Robbie's heartbeat — now a dedicated cron.

Disabled / One-Shot Jobs (Historical)#

These jobs are either disabled (enabled: false), one-shot (deleteAfterRun: true), or have been superseded. Listed for audit completeness.

Name	Agent	Schedule	Notes
`cron-health-opus`	Debbie	Every 30 min (`0,30 * * * *`)	Disabled. Previously compared Opus vs Qwen health check quality. Superseded by fallback-monitor-hourly.
`cron-health-qwen`	Debbie	Every 30 min (`15,45 * * * *`)	Disabled. Counterpart to cron-health-opus, Qwen side.
`t040-review-trigger`	Debbie	One-shot (2026-05-08T14:00Z)	Executed and auto-deleted. Compared Opus vs Qwen after 24h.
`📋 Review task t-092`	Kelly	One-shot (2026-05-08T06:44Z)	Disabled. Auto-review of newly created tasks. Failed (edit permission).
`📋 Review task t-109`	Kelly	One-shot (2026-05-08T10:22Z)	Disabled. Same pattern — task review. Failed (edit permission).
`📋 Review task t-120`	Kelly	One-shot (2026-05-08T10:36Z)	Disabled. Same pattern. Failed (gateway restart during run).
`📋 Review task t-121`	Kelly	One-shot (2026-05-08T10:38Z)	Disabled. Same pattern. Failed (gateway restart during run).
`📋 Review task t-142`	Kelly	One-shot (2026-05-08T12:36Z)	Disabled. Same pattern. Failed (edit permission).
`📋 Review task t-145`	Kelly	One-shot (2026-05-08T12:53Z)	Disabled. Same pattern. Failed (edit permission).
`📋 Review task t-146`	Kelly	One-shot (2026-05-08T13:07Z)	Disabled. Same pattern. Failed (gateway drain during restart).
`📋 Review task t-175`	Kelly	One-shot (2026-05-08T18:01Z)	Disabled. Same pattern. Failed (edit permission).
`discord-task-watch`	Debbie	Every 60s	Disabled. Previously watched Discord channels and created tasks from messages.

🚀 Auto-Dispatch Engine (Every 5min)#

Field	Value
Job ID	`c3c1893f-714b-45d3-8f0b-5ee806fa28ce`
Name	`auto-work-dispatch`
Agent	Debbie 🦑
Schedule	Every 5 minutes (`every 5m`)
Session target	Isolated (light context)
Delivery	None
Enabled	✅ Yes
Success indicator	Calls POST /api/tasks/auto-dispatch; responds with ok=true and dispatch report
Description	Central auto-dispatch engine. Runs every 5min to check backlog for ready tasks, auto-assigns by agent capacity, unblocks dependent tasks, and age-escalates stale backlog items. Updated from 15m → 5m in t-1015. Timeout: 120s, light context.

🔥 Oscar Agent Poll (Every 5min)#

Field	Value
Job ID	`fd838e8e-583b-44ce-be0d-c1aabf0963fc`
Name	`agent-poll-oscar`
Agent	Oscar 🔥
Schedule	Every 5 minutes (`every 5m`)
Session target	Isolated (light context)
Delivery	None
Enabled	✅ Yes
Success indicator	Oscar self-polls backlog; picks up highest-priority task if idle
Description	Oscar self-poll for backlog checking. Every 5min, Oscar checks his task backlog for ready work (monitoring, reporting, trend analysis). Created in t-1015 — Oscar previously had no agent poll. Timeout: 120s, light context.

Schedule Overlap Map (SGT)#

Time-of-day view showing which crons run at the same time.

00:00 ──────────────────────────────────────────────────────────
01:00 ──────────────────────────────────────────────────────────
02:00 ── mission-control-git-sync ── workspace-daily-sync
03:00 ── kelly-daily-docs-review ── task-pipeline-audit
04:00 ──────────────────────────────────────────────────────────
05:00 ──────────────────────────────────────────────────────────
06:00 ──────────────────────────────────────────────────────────
07:00 ──────────────────────────────────────────────────────────
08:00 ── oscar-daily-cost-check
09:00 ── morning-digest
10:00 ── (Sunday only) vera-weekly-quality-deep-dive
       ── (Sunday only) kelly-weekly-index-sweep (at 02:00)
...
22:00 ──────────────────────────────────────────────────────────
23:30 ── daily-reflection-aggregate

Every 5 minutes:
  :00 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar
  :05 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar
  :10 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar ── agent-poll-debbie
  :15 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar ── agent-poll-kelly ── kelly-task-qc-15min ── rate-limit-recovery-probe ── team-health-every-15min
  :20 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar
  :25 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar
  :30 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar ── agent-poll-debbie ── agent-poll-vera
  :35 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar
  :40 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar ── agent-poll-debbie
  :45 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar ── agent-poll-kelly ── kelly-task-qc-15min ── rate-limit-recovery-probe ── team-health-every-15min
  :50 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar
  :55 ── auto-work-dispatch ── agent-poll-chris ── agent-poll-oscar

Hourly crons (on the hour):
  :00 ── fallback-monitor-hourly ── review-flag-scan-1h ── adoption-verifier-hourly  [all staggered 5min]

Every 2 hours:
  Even hours (:00) ── vera-code-review-2h [staggered 15min]

Potential Conflicts#

02:00 SGT — Three jobs fire simultaneously: mission-control-git-sync, workspace-daily-sync, and task-audit-nightly. These are all isolated sessions and non-conflicting but worth monitoring if the gateway becomes overloaded.
Hourly :00 SGT — fallback-monitor-hourly, review-flag-scan-1h, and adoption-verifier-hourly all fire within the same slot. Each has staggerMs: 300000 (5min), so they're offset by ~5 minutes each — not strictly simultaneous but close. If the gateway shows load at :00, consider adjusting stagger windows for wider spacing.

🛡️ First-Run Verifier#

Field	Value
Job ID	`7ec73442-4e13-492b-ac49-9a470fb4110f`
Name	`first-run-verifier`
Agent	Debbie 🦑
Schedule	Every 15 minutes (`/15 * * *` Asia/Singapore)
Session target	Isolated
Delivery	None (alerts written to disk as JSON files)
Enabled	✅ Yes
Description	Snapshot cron registry → detect new jobs → schedule verification task at (first_run + 1h) → check run history when due → alert if 0 runs found. Bootstrap mode ignores jobs created >48h ago on first run to avoid mass-flagging existing jobs.

Script location: ~/.openclaw/workspace/scripts/first-run-verifier.py State file: ~/.openclaw/workspace/state/first-run-verifier-state.json Alert directory: ~/.openclaw/workspace/state/alerts/ Log file: /tmp/first-run-verifier.log

Does NOT verify these jobs (human-created): 8 "📋 Review task t-NNN" jobs + discord-task-watch.

Failure mode: If the script itself fails, the cron will show an error in its run log. The state file tracks last_run_status and the script will retry failed verification scheduling on subsequent runs. If a verification cron detects 0 runs for a job, it writes an alert JSON file to state/alerts/ for pickup by the WhatsApp delivery pipeline.

Success Criteria Summary#

Job	What "success" looks like	Escalation if failure
morning-digest	WhatsApp delivered with crew summary	Missed digest = 📋 Tier 2 failure (Robbie investigates)
kelly-daily-docs-review	Docs updated/confirmed, SHA advanced	Missed review = 📋 Tier 2 (no immediate impact, caught next day)
kelly-weekly-index-sweep	INDEX.md verified	Missed sweep = 📋 Tier 2 (caught next Sunday)
kelly-task-qc-15min	Backlog tasks QCd or none found	Missed sweep = 📋 Tier 2; unprocessed tasks pile up in backlog until next cycle or manual review
mission-control-git-sync	Git push completes	Consecutive 3 failures = Tier 3 (WAKE Maxime per FAILURE-HANDLING.md §2)
workspace-daily-sync	Backup sync status=ok	Script handles escalation (streak ≥3 = WhatsApp alert); Tier 3 if persistent
vera-code-review-2h	New commits reviewed & posted	Missed review = 📋 Tier 2 (no immediate impact)
vera-weekly-quality-deep-dive	Quality report posted to Discord	Missed = 📋 Tier 2
task-pipeline-audit	Script runs cleanly; legacy statuses normalized; warnings flagged; errors escalated	Missed = 📋 Tier 2 (tasks with skipped stages accumulate silently); errors (script itself fails) = Tier 3 (pipeline integrity gap)
oscar-daily-cost-check	Cost data delivered to WhatsApp	Missed = 📋 Tier 2
daily-reflection-aggregate	Summary written + patterns updated	Timeout (as seen) = 📋 Tier 2; check if timeoutSeconds needs adjustment
fallback-monitor-hourly	Alerts delivered, log appended	🔴 Missed rate-limit alerts = Tier 2 (log gap) / Tier 3 if model unavailable
rate-limit-recovery-probe	Recoveries detected & alerted	🔴 Missed recovery = Tier 2; recovered models stay in fallback state unnecessarily
review-flag-scan-1h	Review-flagged tasks reported	Missed = 📋 Tier 2 (low urgency since Maxime can check manually)
auto-work-dispatch	Central dispatch runs every 5min, auto-assigns backlog tasks	Missed = 📋 Tier 2; backlog tasks won't auto-dispatch until next cycle. Consecutive 5 failures = Tier 3 (dispatch is core pipeline).
agent-poll-oscar	Oscar self-polls backlog every 5min, picks up tasks	Missed = 📋 Tier 2; Oscar won't pick up new tasks autonomously. Consecutive 5 failures = Tier 3.
team-health-every-15min	Health checks run every 15min, alerts sent when thresholds breached	Missed = 📋 Tier 2; stale tasks won't be caught until next cycle or manual check. Consecutive 5 failures = Tier 3 (health monitoring is a safety net).
first-run-verifier	New cron jobs detected and verification scheduled; due checks find runs or write alerts	Script failure = 📋 Tier 2 (missing new-cron-verification coverage until fixed). Consecutive 3 failures = Tier 3 (safety net for silent never-fires is down).
adoption-verifier-hourly	Pending adoption checks scanned; reopened tasks alerted to WhatsApp	Missed = 📋 Tier 2; process-design tasks stay 'done' without actual adoption. Consecutive 3 failures = Tier 3 (adoption verification is a quality gate).

Failure Handling: See
- §1.1 — Automatic detection paths (exit codes, timeouts, OOM)
- §2 — Triage tiers: Tier 3 escalation applies when any cron fails 3+ consecutive runs
- §4 — Known Failure Modes registry (FM-001 through FM-007)
- §7 — Daily failure review (Debbie, 02:00 SGT) aggregates cron failures
Reflection Cycle: See
- §4 — Nightly aggregation cron is daily-reflection-aggregate (Debbie, 23:30 SGT)
- §8 — Full cron job specification for the reflection aggregator
- §8.3 — Graceful degradation for the aggregation cron
- §5 — Weekly deep review (Robbie, Friday 16:00) reviews cron performance patterns
Cron Job State: See ~/.openclaw/cron/jobs-state.json
- Per-job runtime data: last run time, status, duration, consecutive errors
- Delivery status tracking
- Used by cron-health monitoring (historical)
Cron Health Log: See
- Historical tracking from the old cron-health-opus/cron-health-qwen experiment
- Deprecated — current health is tracked in job state

Maintenance Notes#

Adding a new cron: Add entry to this registry and update jobs-state.md cross-references. Update the schedule overlap map.
Removing a cron: Move entry to the Disabled/Historical section. Update the registry.
Timeout adjustments: If a cron consistently times out (e.g., daily-reflection-aggregate), update its timeoutSeconds in the job definition.
Audit frequency: Verify this registry against ~/.openclaw/cron/jobs.json every time cron jobs are added or removed. Full audit weekly as part of Kelly's weekly index sweep.
State source of truth: The master job file at ~/.openclaw/cron/jobs.json is authoritative. The workspace copy at ~/.openclaw/workspace/cron/jobs.json is a synced snapshot and may lag.