monitor-patterns

Streams pane output events with the Monitor tool and line-buffered grep patterns.

111|73|Updated Mar 20, 2026
One-click install
npx skills add https://github.com/autopus-ai/autopus-adk --skill monitor-patterns-autopus-ai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: monitor-patterns
Source: https://github.com/autopus-ai/autopus-adk/tree/main/.omp/skills/monitor-patterns
Command: npx skills add https://github.com/autopus-ai/autopus-adk --skill monitor-patterns-autopus-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Waiting for AI provider panes to finish responding usually requires polling loops that burn tokens and add latency. This Skill shows how to use the OMP Monitor tool to stream pane output events in real time, and how to avoid the buffering pitfalls that silently delay or drop those events. ## Core Features & Use Cases - Real-time pane monitoring: Launch one Monitor call per provider pane (claude, codex, antigravity-cli, opencode) and receive each matching stdout line as a notification, with no token cost while waiting. - Line-buffering guardrails: Detect GNU vs BSD grep at runtime and apply the correct line-buffered invocation (grep --line-buffered, stdbuf -oL, or unbuffer) so events are not stuck in pipe buffers. - Timeout and fallback handling: Convert per-provider Monitor timeouts into a 120-second polling fallback, and degrade gracefully to pure polling on non-OMP platforms. - Use Case: During an orchestra Round 2 brainstorm, wait for three AI provider panes to reach their idle prompt, then collect each pane's scrollback exactly once all responses are complete. ## Quick Start Ask the agent to monitor the claude, codex, and antigravity-cli panes with the Monitor tool using line-buffered grep idle patterns and collect each pane's scrollback once all providers finish responding.

Frequently Asked Questions about monitor-patterns

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I monitor a terminal pane for a specific output pattern in real time?▼

Use the Monitor tool with a command that tails the pane output and pipes it through a line-buffered grep matching your pattern. Each matching stdout line arrives as a notification, so no polling loop or token spend is needed while waiting.

Why is my grep in a pipe not producing output until the command finishes?▼

grep defaults to fully-buffered output when writing to a pipe, so lines are held until the buffer flushes. Add --line-buffered on GNU grep, use stdbuf -oL on macOS BSD grep, or wrap the command with unbuffer.

Does grep --line-buffered work on macOS?▼

macOS ships BSD grep, which does not support --line-buffered. Use stdbuf -oL grep instead, or install GNU grep via Homebrew. The skill includes a runtime detection snippet that picks the right invocation automatically.

What happens when the Monitor tool times out waiting for a response?▼

On timeout, only that provider switches to a polling fallback: sleep 120 seconds, then read the pane scrollback directly. Other providers' Monitor streams keep running, and non-OMP environments use the polling path for all providers.

Can one Monitor call watch multiple provider panes at once?▼

No. Multiplexing several panes into one Monitor makes it impossible to tell which provider emitted a notification. Launch one Monitor call per provider pane, each with its own idle-prompt regex pattern.