ui-deep-optimizer
UI Deep Optimizer
Run /ui-deep-optimizer <source-path|app|url|screenshot|spec> to evaluate an existing interface.
Default: repair editable targets; report findings for read-only targets. Accept
--read-only, --context=<brief>, --max-iter=N (1–10; default 5 complete sweeps),
--budget-minutes=N (positive; optional), and --research=<method-ids>.
Flags are instructions to the agent, not an autonomous browser command.
If no target exists, ask for one. Creating this skill needs no application target.
Read references/methods.json through the helper one pass at a time. Read references/standards.md before assigning standards failures; read references/report-format.md before creating the ledger. Examples and evaluation cases are in evals/evals.json. All paths below resolve from this skill folder.