Can I move agents between different scopes?

Yes, the skill supports migrating agents between project-level, user-level, and plugin-level directories while maintaining references.

Does it validate my system prompt changes?

Absolutely. It validates your modifications against structural best practices, ensuring you use proper XML tags and specific workflow instructions.

How does the skill handle security-sensitive changes?

It includes a risk assessment phase that proactively challenges decisions like enabling bypassPermissions or removing tool restrictions.

Does it support bulk editing of agents?

The skill is designed to discover all agents and then guide you through the selection and modification of specific agents sequentially.

Can I use this to improve my agent's auto-invocation?

Yes, it analyzes your agent's description for action-oriented trigger phrases and suggests improvements to make auto-invocation more reliable.

Modify Agent

Name: Modify Agent
Author: filipowm

byfilipowm

0•

開発者ツール

Updates and refactors existing Claude Code subagents across project, user, and plugin scopes with automated analysis and risk assessment.

The modify-agent skill provides a structured workflow for evolving your Claude Code subagents. It automates the discovery of agents across various directories, performs deep structural analysis of frontmatter and system prompts, and identifies improvement opportunities based on prompt engineering best practices. The skill acts as an intelligent assistant that challenges risky modifications—such as broad tool permissions or reduced constraints—while ensuring that updates to workflows and logic are properly formatted using standard XML structures and action-oriented descriptions.

主な機能

01Interactive before-and-after change planning and validation

02Deep analysis of frontmatter fields and system prompt XML structures

03Multi-scope agent discovery across project, user, and plugin levels

04Automated risk assessment for permission and capability changes

050 GitHub stars

06Proactive suggestions for improving agent triggers and workflow specificity

ユースケース

01Refactoring a subagent's system prompt to handle complex new edge cases

02Migrating a local project-level agent to a global user-level scope for broader use

03Optimizing agent descriptions to improve auto-invocation accuracy within Claude Code

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add filipowm/claude-plugins modify-agent

For use in Claude.ai and ChatGPT

主な機能

01Interactive before-and-after change planning and validation

02Deep analysis of frontmatter fields and system prompt XML structures

03Multi-scope agent discovery across project, user, and plugin levels

04Automated risk assessment for permission and capability changes

050 GitHub stars

06Proactive suggestions for improving agent triggers and workflow specificity

ユースケース

01Refactoring a subagent's system prompt to handle complex new edge cases

02Migrating a local project-level agent to a global user-level scope for broader use

03Optimizing agent descriptions to improve auto-invocation accuracy within Claude Code

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add filipowm/claude-plugins modify-agent

For use in Claude.ai and ChatGPT