Severity taxonomy
critical fails production; warning blocks or rewrites at scale; note means "this will work — and then likely regress performance".
Three severities, defined by consequence rather than by feel. Critical: the statement fails outright or takes production down — a constraint that cannot succeed on populated data, a lock that stops writes for the length of a rewrite, an irreversible drop. Warning: it works, but blocks traffic or rewrites data at scale. Note: it works and blocks nothing, but a performance regression is likely.
For the credential family the same three words are read for secrets: critical is a working credential committed to the repository; warning is a name that is very likely a secret, reported so a human decides.
Severity is policy-adjustable, meaning is not
A team on the Team plan can re-rank any rule or silence it. What a severity means never changes, so the block threshold — critical only, warning and above, any finding — reads the same in every repository.
The CLI is open source and needs no account: npx bolvrk check migration.sql
Also in Deterministic rule engine
- Postgres migration safety checks
- Real parse-tree analysis
- Evidence-backed rules
- False-positive discipline
- Fuzz-tested boundaries
- A corpus of 69 rules
- Credential rules: no secret ever reaches the repo
- Secret scanning for any file: bolvrk secrets
- Performance rules: indexing and query health
- Query verification: bolvrk explain