wise-profile

# /wise-profile — session token-budget profile Before asking any user question, read and follow the [question lifecycle](../../references/workflow-host-control.md#keep-asynchronous-questions-open). Keep asynchronous prompts open until answered; this rule does not authorize questions in autonomous or otherwise prompt-free procedures. ## Why this skill exists wise's autonomous skills and workflows default to their standard depth (opus tiers, full research, full reviewer panels). On API billing that depth costs real money; on a subscription it is usually what you want. This skill is the one knob: set the budget once per session, and every profile-sensitive skill reads it silently. What each level means (each consumer carries its exact mapping table; `medium` = the plugin's standard behavior): | level | meaning | |---|---| | `low` | cheapest run that keeps every gate: sonnet tiers for research / implementation (planning/authoring stay on Opus — **Opus 4.8, never Opus 5**: a MUST rule, see below), minimal optional research, solo leads instead of panels, low retry caps | | `medium` | the standard defaults (set this to undo a `low`/`max`) | | `max` | everything on: opus tiers across phases, full research, review findings adversarially verified before apply | The hard invariant, at every level: profiles change model tier / optional-step scope / team size / caps ONLY — never correctness rules. Commit conventions, dirty-tree refusals, review gates, and push refusals are identical...