1 min lesson

You can tune Bugbot review without teaching it bad blocking habits

Describe what "The job: you can tune Bugbot review without teaching it bad blocking habits" changes in practice.

The job: you can tune BugbotCursor's automated PR reviewer that posts inline findings and can push fix commits from isolated VMs. Press Enter for the full definition. review without teaching it bad blocking habits. Start with this check: Rules map to real repository failure modes. Keep it small enough to check the result yourself instead of taking the summary on trust.

Learn more

Full explanation

A scoped BUGBOT.md rule, then depth by risk

Bugbot keeps missing a repo-specific bug class

A scoped BUGBOT.md rule, then depth by risk

SayBugbot reviews our PRs but keeps waving through handlers that read the request body without our shared validator — the exact bug that bit us last month. How do I make it catch that without turning it into a nag?

DoWrite one scoped rule in the repo's BUGBOT.md: flag any API handler that reads req.body without calling the shared validator first.

SeeThe rule lives with the repo, so Bugbot reads it on every review here and nowhere it doesn't apply — the narrowest effective scope, not a global preference.

DoOpen a PR that touches an auth handler and set the review to Deep rather than Quick, because this change is security-sensitive.

SeeQuick is the right depth for small diffs and formatting; Deep costs more but fits complex or security-sensitive code.

SeeBugbot's review now cites the new rule against a concrete line: this handler reads req.body directly, with no validator call.

DoCheck the flagged line against the diff, confirm it's real, and apply the validator fix yourself.

SeeThe rule maps to a real repository failure mode and the finding earned a human-approved fix — you tuned review quality without teaching Bugbot a bad blocking habit.

Learn more

Optional practice