Does it work across different Git repositories?

Yes, it is specifically designed for cross-repo visibility by reading a centralized registry located in the ~/.claude-loops/ directory.

How does Loop Observer track agent status?

It monitors a global registry file and per-repo logs to track start and end events, iteration progress, and process IDs.

What happens if an agent loop crashes?

The skill detects when a tmux session is no longer active without a corresponding end event and generates a synthetic 'crashed' status report.

Can I re-attach to a running loop?

Yes, the skill provides specific tmux attach commands for every running loop it identifies so you can view the live terminal.

What level of detail does it provide for past loops?

It offers full iteration timelines, including durations, exit codes for workers/reviewers, and links to committed configuration summaries.

Agent Loop Observer

Name: Agent Loop Observer
Author: JSai23

byJSai23

•

分析与监控

Monitors and manages agent loops across multiple repositories with real-time status tracking and crash detection.

Agent Loop Observer provides comprehensive observability for AI agent loops, allowing users to track running, completed, and crashed tasks across all local repositories from a single interface. By aggregating data from cross-repo registries, per-repo tracking logs, and Git-committed history, it offers a centralized view of agent activity including iteration counts, stop reasons, and tmux session management. This skill is essential for developers running background AI tasks who need to verify progress, troubleshoot failures, or quickly re-attach to active terminal sessions.

主要功能

01Automatic detection and reporting of crashed loops with synthetic event logging

02Real-time monitoring of running loops with tmux session integration

03Cross-repository visibility of all active and historical agent loops

04Detailed iteration timelines showing worker and reviewer performance metrics

05Advanced filtering by repository, branch, date, or stop reason

061 GitHub stars

使用场景

01Re-attaching to a specific agent's terminal session using provided tmux commands

02Debugging a failed agent loop by reviewing iteration-level logs and stop reasons

03Auditing multiple background AI agent tasks to see which are currently active

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jsai23/claude-tooling loop-observe-a

For use in Claude.ai and ChatGPT

主要功能

01Automatic detection and reporting of crashed loops with synthetic event logging

02Real-time monitoring of running loops with tmux session integration

03Cross-repository visibility of all active and historical agent loops

04Detailed iteration timelines showing worker and reviewer performance metrics

05Advanced filtering by repository, branch, date, or stop reason

061 GitHub stars

使用场景

01Re-attaching to a specific agent's terminal session using provided tmux commands

02Debugging a failed agent loop by reviewing iteration-level logs and stop reasons

03Auditing multiple background AI agent tasks to see which are currently active

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jsai23/claude-tooling loop-observe-a

For use in Claude.ai and ChatGPT