Reference
Line-by-line reviews
What the Critique agent flags — bugs, security, and tech debt — with severity and a one-click fix.
The Critique agent posts findings as inline comments on the exact lines of a pull request, each with a severity and, where possible, a ready-to-apply patch.
What gets flagged
| Category | Examples |
|---|---|
| Correctness | Null/undefined access, off-by-one, wrong async/await, unhandled errors |
| Security | Injection, secret leakage, unsafe deserialization, missing authz checks |
| Tech debt | Duplicated logic, dead code, risky complexity, missing tests on new paths |
Severity
Every finding carries a severity so you can triage fast:
| Severity | Meaning |
|---|---|
| Critical | Ship-blocker — correctness or security defect |
| Warning | Should fix — risk or maintainability |
| Nit | Optional — style/polish (muted by default, see noise control) |
Fix in one click
Where the Fixer agent can produce a safe patch, the finding includes an Apply action that commits the suggestion to the PR branch. You can also ask for one explicitly with @codingassist fix — see Interactive chat.
Related
Was this page helpful?