Skip to content
Get packs

Evidence-Driven Tracing

Explains why an observed result happened by ranking competing hypotheses against evidence

Source Download skill
Repository
Yeachan-Heo/oh-my-claudecode
Stars
40,512
Licence
MIT
Files
1 file

What it does

When installed, the agent stops jumping straight to fixing things when you ask why something happened. Instead it restates what was observed, proposes several different explanations, gathers evidence for and against each, ranks them by how strong the evidence is, and suggests the cheapest test that would tell the leading explanations apart. It is designed to run as a coordinated team of parallel 'tracer' workers inside Claude's built-in team mode, which is a layer on top of the oh-my-claudecode toolkit.

Great for

  • Developers debugging ambiguous regressions, latency problems or unexpected runtime behavior who want a reasoned explanation before changing code
  • Engineers doing postmortems or architecture premortems who need competing causes weighed against evidence
  • Researchers or analysts trying to explain a surprising experimental or benchmark result

Skip it if

  • Simple, well-defined bugs where the fix is already obvious, since the multi-hypothesis process adds overhead
  • Situations needing an immediate code change, because the skill intentionally avoids jumping to implementation
  • Anyone expecting a guaranteed root cause: the skill reports ranked likelihoods and gaps, not certainty

Try asking

  • “/oh-my-claudecode:trace why does our API p95 latency double every Monday morning?”
  • “/oh-my-claudecode:trace the benchmark improved after the refactor but production error rate went up; what are the likely causes?”
  • “/oh-my-claudecode:trace the config change was applied but requests still route to the old service; trace back the likely causes”
Step-by-step guide: how to use Evidence-Driven TracingInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    trace.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip trace.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip trace.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Give a precise observation: state what happened, what you expected, and where you saw it, so the lead can restate it accurately
  • Provide existing logs, traces or benchmark outputs up front, since the skill ranks evidence by strength and primary artifacts rank highest
  • Ask for the critical unknown and discriminating probe first if you want a quick next step rather than a full investigation

What's inside

  • SKILL.md

Redistributed unmodified from Yeachan-Heo/oh-my-claudecode under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.