Subagent Watchdog

A small reliability tool for agentic workflows.

What it does

• After you spawn a subagent with a label, start a watchdog timer.
• The subagent must write a completion marker file.
• If the marker is missing when the timer expires, the watchdog alerts (or exits non-zero so the caller can restart).

Files

• watch.sh — watchdog runner
• SUBAGENT_CONTRACT.md — what the subagent must do

Usage

1) Enforce the contract in your subagent prompt

Tell the subagent to:

• write full results to the instructed project files
• write a marker file: subagent-watchdog/state/.done

2) Start the watchdog

# explicit wait (seconds)
./watch.sh "my-subagent-label" 601

# or omit wait_seconds to read from ~/.openclaw/openclaw.json:
#   caliban.subagentWatchdog.maxRuntimeSeconds (+1 second)
./watch.sh "my-subagent-label"

Notification

By default watch.sh prints a failure message and tries to notify using OpenClaw messaging if:

• OPENCLAW_BIN and WATCHDOG_CHAT_ID are set.

Config read (optional):

• Reads OPENCLAW_CONFIG_PATH or ~/.openclaw/openclaw.json for caliban.subagentWatchdog.maxRuntimeSeconds.

This keeps the skill portable: you can use it with or without OpenClaw messaging.