trace

Generate competing hypotheses and gather parallel evidence to explain observed results.

1|Updated Apr 6, 2026

Search Tags:#debugging #claude #evidence #tracing #hypotheses #team-mode

Ready to run

npx skills add https://github.com/Hyeonjun0527/yeon --skill trace-hyeonjun0527

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: trace
Source: https://github.com/Hyeonjun0527/yeon/tree/main/.codex/skills/trace
Command: npx skills add https://github.com/Hyeonjun0527/yeon --skill trace-hyeonjun0527

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Orchestrates evidence-driven tracing to explain why an observed result happened by generating competing tracer hypotheses and evaluating evidence in parallel.

Core Features & Use Cases

Evidence-driven hypothesis tracing to diagnose ambiguous results
Parallel evidence gathering across multiple hypotheses
Team-mode orchestration with distinct tracer lanes and rebuttal rounds
Clear synthesis with a ranked explanation, critical unknown, and discriminating probe
Applicable to runtime bugs, performance analysis, architecture reviews, and experimental results

Quick Start

Restate the observation exactly, generate three distinct hypotheses, spawn tracer lanes in Claude built-in team mode, and collect evidence for and against each lane to produce a ranked synthesis.

Frequently Asked Questions about trace

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

FAQPage Schema

What is evidence-driven tracing for debugging ambiguous runtime bugs?▼

Evidence-driven tracing is a diagnostic method that explains why an observed result happened by generating competing hypotheses and gathering supporting evidence in parallel. It produces a ranked synthesis to identify root causes.

How do I trace root causes using competing hypotheses in Claude?▼

Restate your observation exactly, generate three distinct hypotheses, and spawn parallel tracer lanes using Claude built-in team mode. Collect evidence for and against each lane to produce a final ranked explanation.

Can I use this tracing approach for performance analysis and architecture reviews?▼

Yes, this tracing approach supports ambiguous, causal, evidence-heavy questions across runtime bugs, performance behavior, architecture reviews, and experimental results by evaluating multiple hypotheses simultaneously.

What's the best way to diagnose causal issues when the root cause is unclear?▼

The best way is to use team-mode orchestration with distinct tracer lanes and rebuttal rounds. This enforces structured evidence gathering for and against multiple hypotheses, yielding a discriminating probe.

Do I need built-in team mode to run parallel evidence gathering?▼

Yes, the workflow requires Claude built-in team mode to spawn distinct tracer lanes, evaluate evidence for and against each hypothesis, and conduct rebuttal rounds for structured synthesis.