deep-dive

Run a two-stage pipeline for root-cause investigation and requirements crystallization.

Updated May 20, 2026
One-click install
npx skills add https://github.com/xdkp/oh-my-claudecode --skill deep-dive-xdkp
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: deep-dive
Source: https://github.com/xdkp/oh-my-claudecode/tree/main/skills/deep-dive
Command: npx skills add https://github.com/xdkp/oh-my-claudecode --skill deep-dive-xdkp

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Deep-dive helps you investigate why something happened and immediately turn the evidence into clear, actionable requirements instead of losing context between separate trace and interview steps.

Core Features & Use Cases

  • Two-stage causal pipeline: runs a 3-lane trace (implementation, config/orchestration, and measurement/assumptions) and then uses the findings for requirements crystallization.
  • 3-point trace injection: enriches the initial problem framing, seeds codebase/system context, and provides an initial question queue derived from unresolved critical unknowns.
  • Resume-ready artifacts: saves trace and final spec outputs to persistent paths for robust continuation.
  • Best-practice handoff: produces a spec that can be forwarded into consensus planning and automated execution.

Quick Start

Use the deep-dive skill when you need to investigate the root cause first, then get a precise plan for what to do next by running: /deep-dive "Why does the production pipeline fail intermittently on the transformation step"

Frequently Asked Questions about deep-dive

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

FAQPage Schema
How do I trace root causes of intermittent bugs and turn them into requirements?

Root cause tracing and requirements crystallization are handled as a two-stage pipeline that runs multi-lane causal tracing and then injects the findings into a targeted interview to produce an actionable spec.

What is the best way to investigate ambiguous production pipeline failures?

Investigating ambiguous failures requires running a 3-lane trace across implementation, config orchestration, and measurement assumptions, then gating progression until ambiguity is low enough for spec generation.

How does multi-lane causal tracing work for complex debugging?

Multi-lane causal tracing works by splitting evidence-based investigation into implementation, config orchestration, and measurement lanes, then seeding codebase context and an interview question queue from unresolved unknowns.

Can I use trace findings to initialize a requirements interview without data injection risks?

Trace findings enrich problem framing and interview initialization with safeguards against treating data as instructions, ensuring evidence context carries forward safely during requirements crystallization.

How do I resume a root cause analysis session after losing context?

Resuming root cause analysis relies on resume-ready artifacts that save trace outputs and final specs to persistent paths, allowing robust continuation of the investigation pipeline.

When should I not use a two-stage trace and interview pipeline?

A two-stage trace and interview pipeline should not be used when a problem is straightforward enough to skip evidence-based investigation or when immediate spec generation is required without ambiguity reduction.