CLI: 'preflight' vs 'provider-preflight' naming is confusing

[rendrag-git]

UX problem

Two subcommands with similar names and overlapping semantics:

• `oc-bench preflight --suite ... --models ...` — benchmark-run preflight (predates M3). Verifies a suite + model config will run end-to-end safely.
• `oc-bench provider-preflight --profile X --provider vllm --base-url Y --route-model Z` — provider-route preflight (M3). Runs four gates against a provider config.

A user reading `--help` has to read both descriptions to know which to invoke. The names don't telegraph the difference.

Options

1. Rename `preflight` → `run-preflight` and `provider-preflight` → `provider-check` (or similar). Breaking change.
2. Subcommand groups: `oc-bench preflight run` and `oc-bench preflight provider`. Cleaner but bigger restructure.
3. Leave as-is, sharpen `--help` text: change the two help strings so the difference is obvious from `oc-bench --help` alone. Lowest effort, lowest payoff.

Recommendation

Option 3 first (1-line fix). Revisit option 2 when the next CLI surface change lands.

Severity

P3 — neither command misbehaves; just easy to invoke the wrong one.

[rendrag-git]

Codex triage: category=enhancement (CLI UX). Impact=P3 (easy to invoke wrong preflight; extra support/reading). Suspected areas: openclaw_bench/cli.py command wiring + help strings for 'preflight' vs 'provider-preflight'. Next action: take the low-risk pass first—tighten the two --help descriptions so the distinction is obvious from 'oc-bench --help' (run-preflight vs provider-route checks); then decide later whether to rename or restructure subcommands.