cron-manager

name	cron-manager
description	Plan, create, edit, remove, list, and test scheduled cron jobs. Each cron job runs as an AI agent session — use the plan→build→create workflow to keep sessions efficient.
user-invocable	false

When to Use

Use this skill for ANY task that needs to happen on a schedule or repeatedly:

• Monitoring / tracking (prices, stocks, website changes, uptime)
• Alerts and notifications ("tell me if X happens")
• Periodic scans (job boards, RSS feeds, news)
• Daily/weekly digests and summaries
• Follow-up reminders
• Recurring data collection
• Anything described as "every N minutes/hours/days" or "check regularly"

Key insight: Each cron run is a full AI agent session — expensive if done frequently. The plan subcommand helps you decide when a backing skill is needed to keep sessions lightweight.

When NOT to Use

NEVER do any of these instead of using this skill:

• Create Python/Node/Bash scripts for monitoring or scheduling
• Use crontab -e or create systemd timers
• Write to ~/.openclaw/cron/jobs.json directly
• Call openclaw cron add, openclaw cron edit, or openclaw cron rm directly
• Create while sleep loops or use watch for recurring tasks
• Install scheduling libraries (APScheduler, node-cron, etc.)

All of the above bypass version control and the sync system. Use this skill instead.

Workflow

ALWAYS start with plan — never jump straight to create.

1. PLAN — Analyze the task

./run.ts plan --task "Track BTC price, alert on 5% drop in 1h" --schedule "*/15 * * * *"

The plan returns:

• tier: script-only (skill does everything), script-agent (skill gathers data, agent interprets), or agent-only (full reasoning needed)
• costWarning: alerts if frequency is expensive
• skillSpec: what skill to create if one is needed (name, interface, description)
• suggestedPrompt: a tools-first prompt template
• workflow: step-by-step instructions to follow

2. BUILD — Create a backing skill (if plan says so)

If skillSpec.needed is true:

skill-scaffold create --name "<skillSpec.suggestedName>" --description "<skillSpec.description>"

Then implement the logic in skills/<name>/run.ts:

• Use direct API calls, not web scraping (CoinGecko, RSS feeds, etc.)
• Output structured JSON to stdout
• Follow the CLI contract: --help, exit codes, JSON output
• Test it: ./skills/<name>/run.ts --help and a real invocation

Add the symlink to module.nix and rebuild.

3. CREATE — Wire it up

./run.ts create --name "BTC Monitor" --schedule "*/15 * * * *" \
  --requires price-monitor \
  --prompt "Run price-monitor check --asset btc --threshold 5 --window 1h. If alert: false, end silently. If alert: true, send details via WhatsApp."

The --requires flag validates that referenced skills exist before creating the YAML.

4. VERIFY

./run.ts list

Usage

./run.ts plan --task "description" --schedule "cron expression"
./run.ts create --name "Job Name" --schedule "0 8 * * *" --prompt "Instructions" [--requires skill1,skill2]
./run.ts edit --name "Job Name" [--schedule "..."] [--prompt "..."] [--session main|isolated]
./run.ts remove --name "Job Name"
./run.ts list [--verbose]
./run.ts test --name "Job Name"
./run.ts --help

Tier Examples

script-only (monitoring):

# Plan identifies: high-frequency + monitoring → script-only, needs price-monitor skill
./run.ts plan --task "Track BTC price, alert on 5% drop" --schedule "*/15 * * * *"
# After building the skill:
./run.ts create --name "BTC Monitor" --schedule "*/15 * * * *" --requires price-monitor \
  --prompt "Run price-monitor check --asset btc --threshold 5 --window 1h. If no alert, end silently. If alert, send via WhatsApp."

script-agent (digest):

# Plan identifies: data gathering + digest → script-agent, use existing job-scanner skill
./run.ts plan --task "Scan job boards and send top matches" --schedule "0 8 * * *"
# job-scanner already exists:
./run.ts create --name "Job Scan AM" --schedule "0 8 * * *" --requires job-scanner,job-tracker \
  --prompt "Run job-scanner --new-only. Filter results for roles matching Enrique's profile (see LORE.md). Add top matches via job-tracker add. Draft top-5 digest. Send to Enrique via WhatsApp. Approval required."

agent-only (reasoning):

# Plan identifies: reasoning-heavy → agent-only, no skill needed
./run.ts plan --task "Weekly self-audit of performance" --schedule "30 10 * * 0"
./run.ts create --name "Self-Audit" --schedule "30 10 * * 0" \
  --prompt "Weekly Self-Audit. Review your performance for the past week..."

Output

JSON to stdout. Each subcommand returns a structured result:

{ "action": "plan", "tier": "script-only", "skillSpec": { "needed": true, "suggestedName": "price-monitor", ... }, "suggestedPrompt": "...", "workflow": [...] }
{ "action": "created", "name": "BTC Monitor", "file": "btc-monitor.yaml", "requires": ["price-monitor"], "syncResult": "..." }

Internals

This skill manages YAML files in /etc/nixos/openclaw/cron/jobs/ and runs cron-sync to register changes with the gateway. You don't need to understand the internals — just use the CLI commands above.