postmortem

Analyze flawed AI agent outputs to identify reasoning failures and propose preventative measures.

46|5|Updated Mar 7, 2026

Search Tags:#debugging #agent behavior #reasoning #error analysis #postmortem #ai safety

Ready to run

npx skills add https://github.com/victoriacity/openakari --skill postmortem-victoriacity

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: postmortem
Source: https://github.com/victoriacity/openakari/tree/main/.claude/skills/postmortem
Command: npx skills add https://github.com/victoriacity/openakari --skill postmortem-victoriacity

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill helps you understand why an AI agent produced a flawed output, focusing on the reasoning process that led to the error and why it wasn't caught.

Core Features & Use Cases

Root Cause Analysis: Identifies the specific reasoning failure mode (e.g., anchoring, context loss, ungrounded generation).
Prevention Strategy: Determines what checks, conventions, or tools could have prevented the flaw.
Use Case: If an agent incorrectly reported a research finding, use this skill to trace back how the error occurred and propose a new verification step to prevent similar mistakes in the future.

Quick Start

Use the postmortem skill to analyze the failure described in the log entry 'Agent produced incorrect summary for document X'.

Frequently Asked Questions about postmortem

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

FAQPage Schema

How do I analyze AI agent reasoning failures after an incorrect output?▼

To analyze AI agent reasoning failures, you examine flawed outputs to identify failure modes like anchoring or context loss, then determine why the error occurred and propose preventative measures to avoid similar mistakes.

What is a postmortem analysis for AI agent behavior?▼

A postmortem analysis for AI agent behavior diagnoses reasoning failures by asking why a specific flaw wasn't caught, classifying the failure mode, and bridging prevention actions to future tasks rather than focusing on technical data errors.

How do I prevent ungrounded generation errors in AI agents?▼

To prevent ungrounded generation errors in AI agents, you perform a root cause analysis to identify the reasoning failure, then determine what verification steps, checks, or conventions could have stopped the flawed output from being accepted.

Can I trace context loss in an AI agent's reasoning process?▼

Yes, you can trace context loss by analyzing the flawed AI agent output to identify the specific reasoning failure mode, focusing on why the context degradation wasn't caught and proposing new verification steps to prevent it.

What's the best way to find root causes of flawed AI summaries?▼

The best way to find root causes of flawed AI summaries is to trace back how the reasoning error occurred, classify the failure mode, and propose a new verification step to prevent similar mistakes in future document processing tasks.