Bug triage¶
This profile does not change code. It writes a short report on each issue as a comment, or as a new issue for the scheduled run. A human changes the issue fields.
Use when¶
- New bug reports arrive and must be sorted.
Do not use when¶
- A bug is already confirmed and has reproduction steps. Use debugging.
Flow¶
- Start the investigator in a fresh session for each new bug report.
- Find duplicates. Search open and closed issues for the same symptom.
- Try to reproduce the bug from the report. Use only a sandbox or a test environment.
- Write the triage report as an issue comment:
- Reproduced: yes, no, or not tried (with the reason)
- Reproduction steps, expected result, and actual result
- Suspected area of the code (file paths), with a confidence level
- Suggested severity and labels, each with a one-sentence reason
- Duplicate of: issue number, or none
- A human approves the labels and the severity (G4). The investigator posts the comment only. A human applies the labels. Label application by a role is not built yet (D51, issue #79).
- If the bug is real, a human marks the issue ready for the debugging profile. This step is not built yet (D51, issue #79).
WARNING: The investigator reads untrusted text from bug reports. Do not give it production data or an outbound channel other than issue comments (P9, lethal trifecta).
Exit criteria¶
- Each new report has a triage comment.
- Each confirmed bug has reproduction steps, and a human approved its severity.
Failure handling¶
| Failure | Action |
|---|---|
| The bug cannot be reproduced | A human labels it "needs info". Ask the reporter for the missing facts. |
| The report has a secret or private data | Do not copy it into the comment. Tell the repo owner. |
Evidence¶
Dossier IDs: F4, H4, D1.