Module rules 

Tier-1 deterministic rule engine.

Classifies a ProposedCommand into Class::Safe, Class::Catastrophic, or Class::Ambiguous using only fixed rules — never a model. This is the security spine: the block decision for catastrophic commands lives here and cannot be argued past.

Design bias: catastrophic checks run first and broadly (a false “this is dangerous” is recoverable; a missed catastrophe is not — see the zero- tolerance rule in CLAUDE.md). Only confidently read-only/build/test commands are marked Safe. Everything else is Ambiguous, to be held or scored.

This module performs no I/O: it reasons purely about the command text, so it is deterministic and trivially testable.

Structs

RuleMatch

The result of classifying a command: its class and the rule that decided it.

Functions

classify

Classify a proposed command. Always returns; never panics.

classify_and_decide

Classify a command and produce a full Tier-1 verdict for the given mode.

classify_line

Classify a raw command line (the entry point used by tests too).

decide

Map a class to a decision for the given mode (Tier-1, rules-only).