| name | debug |
| description | Investigate bugs, errors, or unexpected behaviour — reproduce, isolate, diagnose, and fix. Use when given an error message, stack trace, "works in staging but not prod", or "broke after deploy". Do NOT use for incident response workflows, code review (code-review), or implementation (implement).
|
| license | MIT |
| allowed-tools | Read Write Glob Grep Bash |
| argument-hint | <error message or problem description> |
| metadata | {"author":"Carinya Parc","version":"0.1.0","owner":"web-development","review_cadence":"as-needed","work_shape":"targeted-change","output_class":"code-change"} |
Debug
Structured debugging — reproduce, isolate, diagnose, and fix. Pass the error
message, stack trace, or problem description after the skill name.
See CONNECTORS.md for monitoring and source-control tools.
Steps
- Reproduce — expected vs actual; exact steps; scope (when started, who affected)
- Isolate — narrow component/service/path; check recent deploys, config,
dependencies; review logs and exact error text
- Diagnose — form hypotheses and test them; trace the code path; find root
cause (not just symptoms)
- Fix — propose a fix with explanation; consider side effects and edge cases;
suggest regression tests
What to gather
- Exact error message or stack trace (do not paraphrase)
- Steps to reproduce; expected vs actual; scope
- What changed recently (deploys, deps, config)
- Logs or screenshots
If monitoring is connected: pull logs, error rates, and metrics around the issue
time; correlate with recent deploys. If source control is connected: identify
commits/PRs on affected paths. If project tracker is connected: search related
bugs; create a ticket for the fix once identified.
Output format
## Debug Report: [Issue Summary]
### Reproduction
- **Expected**: …
- **Actual**: …
- **Steps**: …
### Root Cause
…
### Fix
…
### Prevention
- [Test to add]
- [Guard to put in place]