What is the primary purpose of the identify skill?

It is an automated tool designed to scan AI agent session logs for errors, knowledge gaps, and performance issues to facilitate continuous learning.

How does it handle previously processed logs?

It checks the last learning timestamp and skips all entries prior to that date to ensure only new, relevant issues are processed.

What kind of issues does it detect?

It identifies incorrect outputs, knowledge gaps, missed context, poor judgment, and systemic pattern failures within agent transcripts.

Does this skill fix the issues it finds?

No, it focuses strictly on identification and reporting. Separate investigation and incorporation skills handle the root cause analysis and knowledge base updates.

Can I invoke the identify skill manually?

No, this skill is not user-invocable. It is designed to be called by other system agents or workflows within the learning-agents framework.

AI Agent Performance Auditor

Name: AI Agent Performance Auditor
Author: Unsupervisedcom

byUnsupervisedcom

•

데이터 과학 및 ML

Analyzes AI agent session transcripts to identify performance issues, knowledge gaps, and incorrect outputs for continuous improvement.

The identify skill acts as a specialized quality assurance layer for autonomous AI agents within the Deepwork framework. It systematically reviews conversation logs and session transcripts to surface specific instances where an agent failed to meet performance standards—such as providing incorrect code, missing context, or exhibiting poor judgment. By processing raw JSONL logs and filtering for new activity since the last training cycle, it generates structured issue reports that serve as the foundation for agent learning and behavioral refinement. This skill automates the critical feedback loop necessary for maintaining high-reliability AI systems.

주요 기능

0135 GitHub stars

02Delta-based review to skip previously processed timestamps

03Duplicate issue prevention and folder cleanup

04Categorized issue identification (knowledge gaps, poor judgment, etc.)

05Structured reporting for agent training loops

06Automated session transcript analysis

사용 사례

01Identifying systemic knowledge gaps in domain-specific AI models

02Automating the QA feedback loop for autonomous AI agents

03Generating structured error reports from raw agent session logs

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add unsupervisedcom/deepwork identify

For use in Claude.ai and ChatGPT

Download Skill