investigate

Diagnose root causes and inconsistencies across code, performance, and systems.

18|3|Updated Mar 28, 2026

Search Tags:#debugging #troubleshooting #investigate #code-analysis #root-cause #software-engineering

Ready to run

npx skills add https://github.com/ondrej-svec/heart-of-gold-toolkit --skill investigate-ondrej-svec

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: investigate
Source: https://github.com/ondrej-svec/heart-of-gold-toolkit/tree/main/plugins/deep-thought/skills/investigate
Command: npx skills add https://github.com/ondrej-svec/heart-of-gold-toolkit --skill investigate-ondrej-svec

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Detective-style investigation that follows evidence trails to find root causes, bugs, inconsistencies, and hidden problems across code, performance, architecture, data, and systems. Three investigative lenses guide analysis: Sherlock Holmes for deduction, Hercule Poirot for intent, and Columbo for missing pieces.

Core Features & Use Cases

Phase-driven analysis: Establish the case, survey the scene, examine evidence, interview witnesses, and connect the evidence to conclusions.
Multi-perspective reasoning: Apply deduction, psychology/intent, and persistence to surface root causes and edge cases.
Use Case: Diagnose a flaky API by tracing logs, traces, and tests to identify the failing condition and required fix.

Quick Start

Provide a concise, structured investigation plan by loading available evidence and outlining next steps.

Frequently Asked Questions about investigate

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

FAQPage Schema

How do I find the root cause of a flaky API or software bug?▼

A detective-style code investigation diagnoses root causes by following evidence trails through phased analysis—Establish Case, Survey Scene, Examine Evidence, Interview Witnesses, Connect Evidence—to surface bugs and hidden problems across code, performance, architecture, data, and systems.

What is the best way to troubleshoot code inconsistencies across a large software project?▼

The best way to troubleshoot code inconsistencies across a large software project is applying multi-perspective reasoning—deduction, intent, and persistence—to trace evidence trails and connect findings to root causes across any software project size.

How do I start a code investigation when diagnosing hidden software problems?▼

Start diagnosing hidden software problems by loading available evidence and outlining a structured investigation plan that surveys the scene, examines evidence, and connects findings to actionable conclusions.

Can I perform read-only code analysis without modifying the software architecture?▼

Yes, you can perform read-only code analysis without modifying the software architecture by using only provided tools like Read, Grep, Glob, Agent, AskUserQuestion, and Bash to follow evidence trails and diagnose issues.

When should I use a phase-driven investigation workflow for debugging?▼

Use a phase-driven investigation workflow for debugging when you need to follow evidence trails across code, performance, architecture, data, and systems to surface hidden problems and edge cases through multi-perspective reasoning.