Can it prevent the same mistake from happening twice?

Yes, it utilizes a prevention hierarchy that prioritizes automated fixes and hooks over weaker methods like documentation, ensuring more reliable long-term performance.

Why is history verification mandatory in this skill?

It requires using get-history to verify the actual sequence of events, ensuring the root cause analysis is based on objective data rather than the agent's potentially unreliable memory.

What methodologies does it use for error analysis?

It provides three distinct methods: Method A (5-Whys), Method B (Modular Error Taxonomy), and Method C (Causal Barrier Analysis) to ensure a systematic approach to debugging.

How does this skill help with long AI conversations?

It tracks token usage, message counts, and compaction events to determine if context degradation or memory saturation is the root cause of an agent's mistake.

Learn From Mistakes

Name: Learn From Mistakes
Author: cowwoc

bycowwoc

•

Analytics & Monitoring

Analyzes AI agent failures using structured root cause analysis and context monitoring to prevent recurring errors.

This skill provides a rigorous framework for analyzing mistakes within Claude Code sessions, specifically targeting context degradation and token-related failures. By utilizing methodologies like 5-Whys, modular error taxonomy, and causal barrier analysis, it helps developers identify whether a failure was caused by logic, environment, or AI memory saturation. It enforces a mandatory history verification step and a multi-level prevention hierarchy to ensure that identified issues are resolved with robust, automated checks rather than fragile documentation updates.

Key Features

01Mandatory session history verification to eliminate analysis bias

02Structured prevention hierarchy for implementing robust automated fixes

03Automated token and context tracking to identify quality degradation thresholds

04Context degradation checklists for monitoring AI performance over time

052 GitHub stars

06Multi-method Root Cause Analysis including 5-Whys and Causal Barrier Analysis

Use Cases

01Identifying the optimal point for task decomposition when context windows fill up

02Investigating why an agent ignored specific project instructions or constraints

03Debugging logical errors or build failures in long, complex AI sessions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cowwoc/claude-code-cat learn-from-mistakes

For use in Claude.ai and ChatGPT

Key Features

01Mandatory session history verification to eliminate analysis bias

02Structured prevention hierarchy for implementing robust automated fixes

03Automated token and context tracking to identify quality degradation thresholds

04Context degradation checklists for monitoring AI performance over time

052 GitHub stars

06Multi-method Root Cause Analysis including 5-Whys and Causal Barrier Analysis

Use Cases

01Identifying the optimal point for task decomposition when context windows fill up

02Investigating why an agent ignored specific project instructions or constraints

03Debugging logical errors or build failures in long, complex AI sessions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cowwoc/claude-code-cat learn-from-mistakes

For use in Claude.ai and ChatGPT