error-log

Extract structured error reports from AI conversation context and save them as markdown files.

7|1|Updated Apr 1, 2026
One-click install
npx skills add https://github.com/gyy0592/claude-config --skill error-log
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: error-log
Source: https://github.com/gyy0592/claude-config/tree/main/skills/error-log
Command: npx skills add https://github.com/gyy0592/claude-config --skill error-log

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Conversations with AI can hide recurring failure modes because individual mistakes vanish once the exchange ends; this Skill captures each clear mistake as a structured, persistent data point so patterns can be analyzed and systemic fixes applied.

Core Features & Use Cases

  • Trigger detection: activates on explicit /error-log invocation or on clear user frustration (including multi-language examples).
  • Automated diagnosis: extracts timestamp, model, task, category, trigger, expected vs actual behavior, root cause, and severity from conversation context.
  • Review and append-only persistence: presents the generated report for user confirmation and then writes a markdown file to a configurable directory using a strict filename convention.
  • Use case: immediately record a hallucination, destructive action, or skill violation during a development session so teams can audit failures and improve skill templates.

Quick Start

Invoke /error-log or say "log this error" to generate, review, and save a structured markdown error report to the default ~/.claude/error-logs/ directory.

Frequently Asked Questions about error-log

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

FAQPage Schema
How do I log AI mistakes and hallucinations during a conversation?

To log AI mistakes during a conversation, you can invoke /error-log or express clear frustration. The system analyzes context to extract timestamp, model, task, category, trigger, expected vs actual behavior, root cause, and severity, generating a structured markdown report for your confirmation before saving.

What is structured error logging for AI incidents?

Structured error logging for AI incidents captures model mistakes as persistent data points using a six-category classification. It records metadata including ISO timestamps, triggers, root causes, and fixes, preventing individual failures from vanishing when a conversation ends so patterns can be analyzed.

How do I save an incident report for a Claude AI error?

You can save an incident report for a Claude AI error by triggering the logging function, reviewing the automatically extracted diagnosis, and confirming the write. The report is appended to a markdown file in a configurable directory, defaulting to ~/.claude/error-logs/, using a strict filename convention.

Can I automatically capture an audit trail when an AI makes a destructive action?

Yes, you can automatically capture an audit trail when an AI makes a destructive action. The logging mechanism activates on explicit invocation or clear user frustration, extracts the root cause and expected versus actual outcomes, and stores the report append-only to preserve the incident history.

Does error logging work without requiring additional dependencies?

Yes, error logging works without requiring additional dependencies. It operates natively within the conversational session by analyzing existing context and metadata to generate the structured report, requiring only user confirmation before writing the markdown file to the local directory.

What are the limitations of automated AI error diagnosis?

A limitation of automated AI error diagnosis is that it requires user confirmation before writing the report, meaning it cannot save logs fully autonomously. Additionally, it relies on analyzing conversation context, so it only captures mistakes that occur within the current interactive session.