The persistent --quiet/-q and --verbose/-v flags were bound on the root
flag set but read nowhere -- pure no-op flags advertising behavior they
never delivered.
Decision: REMOVE (Option B), not WIRE. A review of every output site
showed there is no non-essential, human-facing output to gate behind
--quiet (the only stdout output is the machine report from audit/check/
apply and essential confirmations like "wrote X"), and no diagnostic or
tracing facility for --verbose to enable (the model's ResolveOptions.
Verbose field is itself dead -- no resolver reads it, nothing writes
diagnostics). Wiring would have required inventing new output purely to
justify the flags, i.e. hollow/cosmetic wiring. Removing the no-ops is
the honest, lower-risk choice for pre-1.0 (v0.3.0).
Passing --quiet/-q/--verbose/-v now yields an unknown-flag usage error
(exit 2). Updated TestNewRootCmd and added TestQuietVerboseFlagsRemoved;
updated README, the design doc, and the CHANGELOG.
The optional LOW item (typed-sentinel refactor of isWriteRefusal) is
deferred: the refusal strings span both internal/applier and
internal/config, so a sentinel refactor would touch the applier
write/refusal path that issue #33 is changing on a separate branch,
risking merge conflicts and cross-package sprawl. Kept this change
scoped to the CLI flag surface.
Closes#34