Trace Investigation Agent FAQs

Question 1

Does the Trace skill modify my source code?

Accepted Answer

No, the Trace subagent is strictly read-only. It uses tools like bash, grep, and glob to investigate and reproduce issues without making any edits to your files.

Question 2

What happens if Trace cannot reproduce the issue?

Accepted Answer

If the failure cannot be reproduced, the agent will report its findings as such, providing any partial evidence gathered and lowering the confidence level of its report.

Question 3

What information is included in a Trace report?

Accepted Answer

A standard report includes the root cause (file and line number), evidence from reproduction, a causal chain of events, recommended corrections, and the agent's confidence level.

Question 4

How does Trace integrate with my workflow?

Accepted Answer

Trace is designed to be the investigative step before fixing. Once it identifies a root cause, the output is typically handed off to a '/fix' command or an orchestrator for resolution.

Question 5

Can Trace work with my existing task management?

Accepted Answer

Yes, if a PG task exists, Trace automatically logs investigation starts, findings, and failures as task comments to keep your project history actionable.

Trace Investigation Agent

主要功能

使用场景

Trace Investigation Agent

主要功能

使用场景