agent-supervisor

Evaluate Claude Code session records and generate global discipline rules.

1|Updated Mar 24, 2026

Search Tags:#ai-agent #claude-code #agent-conduct #behavior-evaluation #discipline-rules

Ready to run

npx skills add https://github.com/choms0521/my-claude-skills --skill agent-supervisor-choms0521

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: agent-supervisor
Source: https://github.com/choms0521/my-claude-skills/tree/main/generated/claude/skills/agent-supervisor
Command: npx skills add https://github.com/choms0521/my-claude-skills --skill agent-supervisor-choms0521

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and agents (resource) and references (resource) components.

What problem does it solve?

This Skill evaluates the behavior of AI agents used in Claude Code, generating and merging global discipline rules based on their performance.

Core Features & Use Cases

Behavior Evaluation: Evaluates agent conduct using a custom judge subagent, analyzing recent Claude Code sessions.
Global Discipline Rules: Generates and merges rules to improve agent behavior across all sessions.
Use Case: For instance, if an AI agent consistently provides incorrect information, this Skill can identify the issue and propose a rule to prevent similar errors in the future.

Quick Start

Run the /agent-supervisor command to evaluate agent behavior and generate discipline rules.

Frequently Asked Questions about agent-supervisor

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

FAQPage Schema

How do I evaluate AI agent behavior in Claude Code?▼

To evaluate AI agent behavior in Claude Code, this Skill uses a custom judge subagent to analyze session JSONL records and assess agent conduct based on recent performance.

How do I generate discipline rules for AI agents?▼

Generating discipline rules for AI agents involves analyzing Claude Code session records to identify behavioral issues and automatically merging new global rules to prevent future errors.

What data do I need to analyze Claude Code agent conduct?▼

You need Claude Code session JSONL records to analyze agent conduct. The Skill processes these session logs to evaluate behavior and generate appropriate discipline rules.

Can I prevent recurring errors from AI agents using behavioral evaluation?▼

Yes, behavioral evaluation identifies recurring issues like consistently incorrect information and proposes global discipline rules to prevent similar AI agent errors in future sessions.

Does agent behavior evaluation require a custom subagent?▼

Yes, agent behavior evaluation requires a custom judge subagent that handles the analysis of Claude Code session records and assesses the conduct of the AI agents.