Dual-host: Claude Code and Codex both write the facts this dashboard needs, but not the same facts. Claude Code gives cost and a hardcoded 5-hour window; Codex gives rate_limits.primary/secondary — a used-percent and a reset instant with a dynamic, host-reported window (observed as 10080 minutes / 7 days, never assumed to be 5 hours) — and no per-token pricing at all. Detect the host you are running under and act accordingly:

Claude Code: run the full pipeline below unmodified (cost, 5-hour window, AI insights, 23-locale HTML).
Codex: run node ${PLUGIN_ROOT}/scripts/run-usage-view.js --gen-agent-prompt --host codex [parsed flags]. It uses the same default AI-analysis pipeline and the same explicit ai / no ai options as Claude Code. This renders the same dashboard (skills/usage-view/template.html via build-report.js) — token breakdown, calendar, session detail — not a separate report. Cost figures read N/A, not $0 or a Claude-derived estimate, and the calendar follows Codex's reported rate-limit lane instead of assuming a fixed 5 hours.
A dual-host skill must not hardcode one host's plugin root. Claude Code exports CLAUDE_PLUGIN_ROOT; Codex does not reliably export CODEX_PLUGIN_ROOT. Resolve PLUGIN_ROOT="${CLAUDE_PLUGIN_ROOT:-${CODEX_PLUGIN_ROOT}}", falling back to this skill's own directory if both are unset.

Parse user arguments, then launch a single background Agent that runs the entire pipeline (analyze → AI insights unless no ai was selected → build → open browser). The user can continue working while the dashboard is being generated.

Help

ONLY show help if the user's argument literally contains the word "help" (e.g. /usage-view help). If no argument or any other argument is given, SKIP this section entirely and proceed to launching the agent.

If the user provides "help" as argument, show usage summary and stop:

/usage-view — Interactive HTML usage dashboard