CRITICAL

The affected output cannot be trusted, or the current execution cannot reliably continue.

Execution examples: corrupt required output, empty selected compile database, malformed metrics required by a report, or unresolved required application symbols.

Content problems should rarely use this level directly. Ordinary model issues, including wrong or missing ASIL values, are reported as HIGH and then gated by the configured failAt threshold. Use CRITICAL for generated evidence or task outputs that are unusable, contradictory, or corrupt.