Summary

Cron jobs with \payload.kind: agentTurn\ have a hardcoded timeout of 120 seconds. Complex agent tasks (e.g. log scanning with multiple tool calls, multi-step analysis) regularly exceed this and get killed mid-execution.

Impact

• Agent tasks that need 3-5 tool calls in sequence (common for monitoring/review jobs) frequently timeout
• Timed-out jobs leave partial state and generate follow-up errors
• No way to configure timeout per-job — all jobs share the same 120s ceiling
• The \ imeoutSeconds\ field in the job payload schema exists but does not appear to be respected for the agent execution phase

Reproduction

1. Create a cron \�gentTurn\ job that performs 4+ sequential tool calls (e.g. read file, exec command, search, write file)
2. Set \ imeoutSeconds: 300\ in the payload
3. Observe that the job still gets killed at ~120s

Expected Behaviour

• Per-job \ imeoutSeconds\ should be respected
• Or at minimum, the default should be higher (300s) with a configurable cap

Environment

• OpenClaw 2026.2.14
• Windows 10
• Multiple cron agentTurn jobs, several regularly hitting the 120s wall