trace

Orchestrate parallel hypothesis tracing and evidence collection for causal failure analysis.

1|Updated Apr 1, 2026
One-click install
npx skills add https://github.com/OliverOuyang/shuhe-work-skills --skill trace-oliverouyang
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: trace
Source: https://github.com/OliverOuyang/shuhe-work-skills/tree/main/skills/trace
Command: npx skills add https://github.com/OliverOuyang/shuhe-work-skills --skill trace-oliverouyang

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps teams investigate ambiguous, causal, and evidence-heavy incidents by structuring hypothesis generation, parallel evidence collection, and principled falsification so the root cause can be identified or meaningfully narrowed.

Core Features & Use Cases

  • Hypothesis orchestration: Restates the observation, generates multiple competing explanations, and assigns one tracer lane per hypothesis.
  • Parallel evidence gathering: Directs tracer workers to collect evidence for and against each hypothesis, rank evidence strength, and name critical unknowns.
  • Rebuttal and synthesis: Runs a rebuttal round between top explanations, detects convergence, and returns a ranked synthesis with the single best discriminating probe.
  • Use Cases: Runtime regressions, performance investigations, architecture postmortems, experimental result tracing, and config/orchestration behavior analysis.

Quick Start

Trace why the service latency spiked for endpoint /api/payments between 14:00 and 14:10 by restating the observation, generating three competing hypotheses, and instructing team-mode tracer lanes to gather evidence and propose the discriminating probe.

Frequently Asked Questions about trace

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

FAQPage Schema
How do I debug ambiguous runtime failures using evidence-driven tracing?

Evidence-driven debugging traces ambiguous runtime failures by orchestrating parallel hypothesis generation and evidence collection. It assigns independent tracer lanes to gather supporting and contradicting evidence, ranks hypotheses by strength, and proposes a single discriminating probe.

What is the best way to investigate a performance regression with multiple competing hypotheses?

Investigating a performance regression requires generating multiple competing explanations and running parallel tracer lanes to collect evidence from logs, metrics, and configs. The process ranks hypotheses by evidence strength, runs rebuttal rounds, and identifies the root cause through falsification checks.

Can I use hypothesis-driven incident analysis for architecture postmortems?

Hypothesis-driven incident analysis supports architecture postmortems by restating observations, generating competing explanations, and directing team-mode tracer workers to gather evidence. It ranks evidence, detects convergence between explanations, and returns a synthesis with the single best discriminating probe.

How do I trace experimental result discrepancies using logs and metrics?

Tracing experimental result discrepancies uses parallel evidence gathering across logs, metrics, configs, and code artifacts. Tracer lanes collect evidence for and against each hypothesis, identify critical unknowns, execute falsification checks, and output a ranked synthesis pinpointing the causal failure.

Does parallel hypothesis tracing work without external dependencies?

Parallel hypothesis tracing operates without external dependencies, using internal tracer workers to collect evidence and rank explanations. It orchestrates independent lanes that gather supporting and contradicting evidence, run rebuttals, and synthesize results without requiring additional components.

When should I not use falsification checks for debugging causal failures?

Falsification checks may be unnecessary for simple, deterministic bugs where the root cause is already obvious. This structured approach is designed for ambiguous, evidence-heavy causal failures where multiple competing hypotheses require parallel evidence collection and ranking to identify the discriminating probe.