introspect

Diagnose agent loops and generate a minimal recovery plan.

161|21|Updated Mar 26, 2026
One-click install
npx skills add https://github.com/softspark/ai-toolkit --skill introspect-softspark
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: introspect
Source: https://github.com/softspark/ai-toolkit/tree/main/app/skills/introspect
Command: npx skills add https://github.com/softspark/ai-toolkit --skill introspect-softspark

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Agent self-debugging and recovery. Use when the agent is stuck in loops, producing repeated errors, or when output quality degrades.

Core Features & Use Cases

  • Structured failure capture: records the last goal, actions, and errors for precise debugging.
  • Pattern classification: maps symptoms to root causes and selects a focused recovery.
  • Recovery action guidance: prescribes the smallest safe fix to regain progress.
  • Self-correction protocol: outputs an introspection report with a standardized format for review.

Quick Start

Describe the failure briefly and list the last 3–5 actions, then invoke introspection to generate a recovery plan.

Frequently Asked Questions about introspect

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

FAQPage Schema
How do I debug an AI agent stuck in a loop or producing degraded output?

To debug an AI agent stuck in a loop, you capture the last goal, actions, and errors, then classify failure patterns to perform root-cause analysis and apply a minimal recovery action with safe guardrails.

What is the best way to analyze root causes for repeated errors in long-running agent tasks?

Root-cause analysis for long-running agent tasks is best handled by mapping classified symptoms to specific failure patterns, which generates a focused recovery plan and a standardized introspection report for review.

How do I start self-debugging an AI agent that is failing in iterative scenarios?

Start self-debugging an AI agent by briefly describing the failure and listing the last three to five actions taken, then invoke introspection to generate a structured recovery plan with safe guardrails.

Can I use structured failure capture for self-correction in failure-prone agent workflows?

Yes, structured failure capture records the last goal, actions, and errors for failure-prone workflows, enabling a self-correction protocol that outputs a standardized introspection report to guide recovery.

What are the limitations of applying minimal recovery actions to complex AI agent failures?

The limitation of minimal recovery actions is that they prescribe only the smallest safe fix to regain progress, which may not resolve deeply complex root causes beyond the captured failure patterns and classified symptoms.