What files does this skill need access to?

The skill requires access to the .claude/agent-memory and .claude/skills directories to analyze patterns and check for existing skill duplicates.

How do I ensure the best results from a Retro report?

Maintain the 'Repeated Patterns', 'Pain Points', and 'Lessons Learned' sections in your agents' MEMORY.md files as per the standardized convention.

How does Retro identify new skill opportunities?

It scans the 'Repeated Patterns' section of your agent's memory files to find manual tasks performed multiple times that are suitable for automation.

Why does Retro flag memory files over 200 lines?

Large memory files contribute to context bloat, which can lead to slower response times, higher token costs, and reduced accuracy in AI behavior.

Does Retro create new agents automatically?

No, it provides a detailed proposal for new sub-agents, including their suggested roles, tools, and scope, allowing you to review and implement them as needed.

Claude Agent Retrospective

Name: Claude Agent Retrospective
Author: WHITS-ISLAND

byWHITS-ISLAND

•

生産性とワークフロー

Optimizes AI workflows by analyzing agent memory to suggest skill automation, sub-agent expansion, and context management.

The Retro skill acts as a performance auditor for your Claude Code environment by scanning agent memory files to identify operational inefficiencies. It automatically detects repetitive manual patterns that could be codified into new skills, identifies pain points that justify the creation of specialized sub-agents, and monitors context bloat to prevent performance degradation. By transforming raw interaction logs into structured actionable reports, it helps developers maintain a lean, high-performing multi-agent ecosystem for complex projects.

主な機能

011 GitHub stars

02Analyzes agent memory files for repetitive task patterns and automation opportunities.

03Cross-references interaction logs against existing skills to prevent redundancy.

04Monitors context file size to prevent memory bloat and performance lag.

05Generates structured reports with clear action items for environment optimization.

06Identifies specific pain points to suggest specialized sub-agent expansion.

ユースケース

01Scaling complex projects by determining when to delegate tasks to new specialized sub-agents.

02Managing context window limits by identifying and pruning overgrown agent memory files.

03Identifying manual tasks that should be automated into custom Claude skills.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add whits-island/inspirehub-mobile retro

For use in Claude.ai and ChatGPT

Download Skill