background-agent-pings
npx machina-cli add skill parcadei/Continuous-Claude-v3/background-agent-pings --openclawBackground Agent Pings
Trust system reminders as agent progress notifications. Don't poll.
Pattern
When you launch a background agent, continue working on other tasks. The system will notify you via reminders when:
- Agent makes progress:
Agent <id> progress: X new tools used, Y new tokens - Agent writes output file (check the path you specified)
DO
1. Task(run_in_background=true, prompt="... Output to: .claude/cache/agents/<type>/output.md")
2. Continue with next task immediately
3. When system reminder shows agent activity, check if output file exists
4. Read output file only when agent signals completion
DON'T
# BAD: Polling wastes tokens and time
Task(run_in_background=true)
Bash("sleep 5 && ls ...") # polling
Bash("tail /tmp/claude/.../tasks/<id>.output") # polling
TaskOutput(task_id="...") # floods context
Why This Matters
- Polling burns tokens on repeated checks
TaskOutputfloods main context with full agent transcript- System reminders are free - they're pushed to you automatically
- Continue productive work while waiting
Source
- This session: Realized polling for agent output wasted time when system reminders already provide progress updates
Source
git clone https://github.com/parcadei/Continuous-Claude-v3/blob/main/.claude/skills/background-agent-pings/SKILL.mdView on GitHub Overview
Background Agent Pings uses system reminders as progress notifications for background agents. It lets you continue working on other tasks while the system updates you, avoiding token-wasting polling and reducing context floods. Read agent output only when the system signals completion.
How This Skill Works
You launch a background task with run_in_background=true and specify an Output path like .claude/cache/agents/<type>/output.md. The agent keeps processing while you work on other tasks. Progress reminders (e.g., Agent <id> progress: X new tools used, Y new tokens) appear, and you read the output file only when the system indicates completion.
When to Use It
- You want to run a long-running agent without blocking your workflow.
- You need to avoid token waste from active polling and context flooding from full transcripts.
- You rely on system reminders to surface progress updates instead of manual checks.
- You intend to verify results by reading a single output file after completion.
- You manage multiple agents in parallel and want non-intrusive progress signals.
Quick Start
- Step 1: Task(run_in_background=true, prompt="... Output to: .claude/cache/agents/<type>/output.md")
- Step 2: Continue with next task immediately
- Step 3: When system reminder shows agent activity, check if output file exists and read only upon completion
Best Practices
- Use the prescribed Output path pattern: .claude/cache/agents/<type>/output.md for result writes.
- Launch tasks with run_in_background=true and immediately proceed to other work.
- Monitor progress via system reminders rather than polling for file changes.
- Read the output file only after a completion signal from reminders.
- Avoid using polling tools or TaskOutput floods that clutter the main context.
Example Use Cases
- Launch a data processing agent in the background; continue modeling while it runs, and read output.md only after a completion reminder.
- Run a long-running analysis agent with progress updates like 'X tools used, Y tokens' and fetch results from the output file when signaled.
- Coordinate two agents performing extraction tasks; rely on individual progress reminders instead of polling their logs.
- An agent writes results to .claude/cache/agents/<type>/output.md; you check this file only when the system indicates completion.
- Avoid tailing logs or polling sleeps; use system reminders to decide when to read the final output.
