Severity Policy
How Karandash assigns and uses severity levels for findings.
Severity Levels
| Level | Meaning | Default Behavior |
|---|---|---|
ERROR |
Critical violation that must be fixed | Blocks merge (exit code 1) |
WARNING |
Issue that should be addressed | Reported but does not block |
INFO |
Informational finding | Reported only, never blocks |
How Severity is Determined
- Rule default -- Each rule has a built-in default severity (typically
ERRORfor blocking rules likeconflict,WARNINGfor advisory rules) - Profile override -- Profiles can set a default severity for all rules:
severity: errororseverity: warning - Per-rule override -- Individual rules can override severity in config:
rules:
title-format:
pattern: "^(feat|fix):"
severity: warning # override to non-blocking
Exit Codes
| Code | Meaning |
|---|---|
0 |
All rules passed (no ERROR findings) |
1 |
One or more ERROR findings |
2 |
Configuration error |
Dry Run
The --dry-run CLI flag only validates the config; it runs no rules. There is no --dry-run-scope flag.
The top-level dry-run: true config key sets the is_dry_run condition, which policies can use (for example to
lower severities), and suppresses work per dry-run-scope:
dry-run-scope: ai(also when unset): AI rules make no LLM calls; other rules are enforced normally.dry-run-scope: all: no LLM calls, and nothing leaves the process: platform comments, Slack, webhook, email, third-party outputs and auto-approve are skipped. Console and file outputs still run.
See YAML config: Dry run.