monitoring-the-operator

Live-monitor an operator's Claude Code transcript JSONL for audit-worthy events.

4|Updated Jan 29, 2026
One-click install
npx skills add https://github.com/danmestas/wardrobe --skill monitoring-the-operator
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: monitoring-the-operator
Source: https://github.com/danmestas/wardrobe/tree/main/skills/monitoring-the-operator
Command: npx skills add https://github.com/danmestas/wardrobe --skill monitoring-the-operator

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Live-monitor an active operator session from a worker pane (typically a stasi/wait-watch worker auditing the parent that spawned it). Use when a worker should watch the operator's transcript JSONL in real time, surface audit-worthy events without polling, distinguish active autonomous runs from hangs, and avoid the chat-Stop feedback loop that wakes the operator on every routine eval. Triggers when the user spawns a wait-watch worker and asks it to "watch what I do", "monitor my session", "audit the orchestrator live", "tell me if I do anything wrong", or any equivalent.

Core Features & Use Cases

  • Live monitoring of an operator's active Claude Code session via a worker pane
  • Real-time auditing of the operator's transcript JSONL with minimal context overhead
  • Distinguish active autonomous runs from hangs and avoid triggering Stop loops
  • Triggered when a user spawns a wait-watch worker and asks it to monitor or audit
  • Supports configurable alerts and traceability for post-hoc reviews

Quick Start

Spawn a wait-watch worker and direct it to monitor the operator's transcript in real time.

Frequently Asked Questions about monitoring-the-operator

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

FAQPage Schema
How do I live-audit an active Claude Code operator session in real time?

To live-audit an operator session, spawn a wait-watch worker pane and direct it to monitor the parent operator's transcript JSONL. This surfaces audit-worthy events in real time without polling or waking the operator.

What is the best way to monitor JSONL transcript files without triggering chat-stop loops?

Monitoring JSONL transcripts without triggering chat-stop loops requires a wait-watch worker that avoids routine eval feedback. This approach distinguishes active autonomous runs from hangs while maintaining silent observation of the session.

Can I use a worker pane to distinguish active autonomous runs from hangs in a transcript?

Yes, a worker pane can distinguish active autonomous runs from hangs by anchoring to the operator's transcript JSONL. It applies real-time streaming with explicit operational triggers to identify when the session stalls.

How do I set up real-time streaming alerts for an operator's transcript without polling?

Real-time streaming alerts without polling are achieved by spawning a persistent monitor worker that anchors safely to the transcript JSONL. The monitor uses explicit operational triggers to surface structured alerting events.

Why does monitoring an operator session wake the orchestrator on every routine eval?

Monitoring wakes the orchestrator due to chat-stop feedback loops triggered by routine polling. Using a wait-watch worker with minimal context overhead avoids waking the operator on every eval while still surfacing audit-worthy events.

Do I need to spawn a wait-watch worker to audit the orchestrator live?

Yes, spawning a wait-watch worker is required to audit the orchestrator live. The worker pane watches the parent operator's transcript JSONL, providing traceability for post-hoc reviews and configurable alerts.