trace

Generate competing hypotheses and collect evidence using Claude Code team mode.

Updated Mar 31, 2026

Search Tags:#orchestration #trace #evidence #claude-code #root-cause #hypotheses #team-mode

Ready to run

npx skills add https://github.com/ClementATH/oh-my-claudecode --skill trace-clementath

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: trace
Source: https://github.com/ClementATH/oh-my-claudecode/tree/main/skills/trace
Command: npx skills add https://github.com/ClementATH/oh-my-claudecode --skill trace-clementath

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Ambiguity and conflicting signals in Claude Code runs require a structured approach to determine why results occur; this skill provides an orchestration layer that reasons with competing explanations and gathers evidence in parallel.

Core Features & Use Cases

Restate the observation and generate multiple, deliberately different hypotheses.
Spawn Claude's built-in team-mode lanes to collect evidence for and against each hypothesis.
Rank explanations by evidence strength and output a discriminating probe to resolve uncertainty quickly.

Quick Start

Provide an evidence-driven trace by restating the observation, generating hypotheses, and collecting supporting and opposing evidence using Claude's built-in team mode.

Frequently Asked Questions about trace

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

FAQPage Schema

How do I find the root cause of ambiguous runtime bugs in Claude Code?▼

Root-cause analysis for ambiguous runtime bugs uses evidence-driven tracing to generate competing hypotheses, collect supporting and opposing evidence in parallel via team mode, and rank explanations by evidence strength.

How does evidence-driven tracing evaluate multiple hypotheses in parallel?▼

Evidence-driven tracing evaluates hypotheses by spawning Claude's built-in team-mode lanes to gather evidence for and against each explanation simultaneously, then ranks them by evidence strength to identify the most likely root cause.

Can I use team mode to investigate performance questions and configuration behavior?▼

Yes, team mode applies to performance questions, configuration behavior, architecture decisions, and experiments where multiple hypotheses must be evaluated in parallel to resolve conflicting signals.

What's the best way to resolve conflicting signals from a Claude Code run?▼

Resolving conflicting signals involves restating the observation, generating deliberately different hypotheses, collecting evidence through team-mode orchestration, and outputting a discriminating probe to quickly resolve uncertainty.

Does Claude Code team mode require any external dependencies to perform root-cause tracing?▼

No external dependencies are required; root-cause tracing operates entirely within Claude Code's built-in team mode, using its orchestration rules to generate hypotheses and collect evidence without additional packages.

When should I not use parallel hypothesis tracing for debugging?▼

Parallel hypothesis tracing is not suited for single-cause deterministic errors where the root cause is already obvious, as the overhead of spawning team-mode lanes adds complexity without benefit.