feat(cursor-agent): improve AI agent automation with -p mode and smart wait

[dongsjoa-byte]

Summary

Improve the cursor-agent skill for AI agent automation scenarios.

Changes

• Recommend -p (print) mode for cleaner non-interactive output
• Smart completion detection by checking shell prompt instead of fixed sleep
• Add cursor-run.sh helper script for convenient automation
• Keep interactive mode as alternative for complex tasks
• Add detailed explanation of why -p mode is better for automation

Why -p mode is better

1. Clean output - No TUI, just text/JSON
2. Auto-exit - Exits when task completes
3. Predictable - Works with --output-format text/json

Smart wait mechanism

Instead of:

sleep 60  # Fixed wait, may be too short or too long

Use:

# Check for shell prompt to detect completion
while true; do
  last_line=$(tmux capture-pane -t cursor -p | grep -v '^$' | tail -1)
  if echo "$last_line" | grep -qE '(\$|%)\s*$'; then
    break
  fi
  sleep 3
done

Testing

Tested on macOS with OpenClaw agent:

• ✅ agent -p 'simple query' - works, exits cleanly
• ✅ Smart wait detects completion correctly
• ✅ Timeout protection works

/cc @swiftlysingh (original author)

[openclaw-barnacle]

Thanks for the pull request! This repository is read-only and is automatically synced from https://clawhub.ai, so we can’t accept changes here. Please make updates on the website instead.