What happens if a previous fix doesn't work?

The skill performs an effectiveness check; if a mistake recurs after a fix, it automatically generates an 'Escalation' report to trigger a more robust solution.

Can I manually trigger a retrospective analysis?

Yes, you can manually invoke the analysis at any time using the /cat:run-retrospective command within the Claude Code interface.

How does the skill know when to run a retrospective?

It uses configurable triggers in retrospectives.json, typically firing every 14 days or after 10 mistakes have been recorded by the system.

What kind of action items does this skill generate?

It suggests specific improvements such as adding new PreToolUse hooks, updating documentation, or modifying source code to prevent identified failure patterns.

Retrospective Analysis & Optimization

Name: Retrospective Analysis & Optimization
Author: cowwoc

bycowwoc

•

Productivity & Workflow

Analyzes historical mistake patterns and evaluates the effectiveness of corrective actions to improve AI agent reliability.

About

The Run Retrospective skill implements a robust continuous improvement loop for AI development workflows. By analyzing accumulated mistakes logged by the learn-from-mistakes plugin, it identifies recurring patterns, groups failures by category, and performs a data-driven effectiveness check on previous fixes. This skill is essential for maintaining high-performance AI agents, as it automatically escalates recurring issues and derives prioritized action items to ensure the development environment becomes progressively more stable and error-resistant.

Key Features

Category-based analysis of protocol, build, and git failures
Quantitative effectiveness tracking for previously implemented fixes
2 GitHub stars
Configurable threshold triggers based on time or mistake count
Multi-level escalation management for persistent issues
Automated pattern identification across historical failure logs

Use Cases

Diagnosing persistent protocol violations in complex CI/CD environments
Automating post-mortem analysis after project milestones
Optimizing AI agent behavior through data-driven defensive programming

About

Key Features

Category-based analysis of protocol, build, and git failures
Quantitative effectiveness tracking for previously implemented fixes
2 GitHub stars
Configurable threshold triggers based on time or mistake count
Multi-level escalation management for persistent issues
Automated pattern identification across historical failure logs

Use Cases

Diagnosing persistent protocol violations in complex CI/CD environments
Automating post-mortem analysis after project milestones
Optimizing AI agent behavior through data-driven defensive programming