openclaw - 💡(How to fix) Fix Feature Request: Cron job lifecycle hooks (preCreate / postCreate / onFailure) [1 participants]

Summary

Add configurable lifecycle hooks for cron job operations, enabling teams to enforce creation standards, auto-register jobs in tracking systems, and handle failures programmatically.

Motivation

We manage 60 cron jobs across 19 agents. After a series of reliability issues, we built a hardening standard (cron-ops skill) that requires every job to have:

• A backing script (not pure LLM generation)
• An explicitly pinned model
• A properly sized timeout
• Failure alerting
• Registration in a manifest file

The problem: Skills are guidance, not gates. The agent can (and did) bypass the standard by calling openclaw cron add directly. We caught it the same day we built the standard — the agent created a new job and skipped 2 of 8 required steps.

Currently the only enforcement is a daily audit script that catches violations after the fact. A pre-create hook would catch them before the job goes live.

Proposed Feature

A hooks configuration under cron in openclaw.json:

{
  "cron": {
    "hooks": {
      "preCreate": "python3 /path/to/validate-cron.py",
      "postCreate": "python3 /path/to/register-manifest.py",
      "onFailure": "python3 /path/to/handle-failure.py"
    }
  }
}

Hook Behavior

preCreate:

• Runs before cron add creates the job
• Receives the proposed job config as JSON on stdin
• If exit code ≠ 0, creation is blocked
• stderr is returned to the agent as the rejection reason
• Example: validate that a backing script exists, model is pinned, timeout is set

postCreate:

• Runs after successful job creation
• Receives the created job config (with ID) as JSON on stdin
• Non-blocking — failures are logged but don't roll back the job
• Example: auto-register in a manifest file, set up failure alerting

onFailure:

• Runs when a job hits N consecutive failures (configurable threshold)
• Receives the job config + failure details as JSON on stdin
• Example: auto-disable the job, send an escalation alert, open a GitHub issue

Example preCreate Validator

#!/usr/bin/env python3
import json, sys, os

job = json.load(sys.stdin)
payload = job.get("payload", {})
errors = []

# Require explicit model
if not payload.get("model"):
    errors.append("Model must be explicitly pinned (no default)")

# Require timeout
if not payload.get("timeoutSeconds"):
    errors.append("Timeout must be configured")

# Require backing script for agentTurn jobs
if payload.get("kind") == "agentTurn":
    msg = payload.get("message", "")
    if "python3" not in msg and ".py" not in msg:
        errors.append("Job must reference a backing script")

if errors:
    print("\n".join(errors), file=sys.stderr)
    sys.exit(1)

Additional Context

• We currently work around this with a wrapper script (ops/cron-create.py) but agents can bypass it
• The cron add CLI doesn't support --failure-alert flags — we have to create then immediately cron edit, which is another gap where enforcement falls through
• Our full audit tooling and standards are open source: jayh4wk/openclaw-skills

Impact

This would benefit any team running OpenClaw at scale with multiple agents creating cron jobs. The hook pattern is unopinionated — it doesn't force any specific standard, just provides the enforcement point.

Submitted by (@jayh4wk)