efficient-dispatchlisted

Model-routing orchestration for any expensive frontier model (Fable, Opus, GPT-5.x) - the main model keeps judgment and Q&A review, mechanical subagent dispatches carry an explicit cheaper model, and ax measures whether the routing actually worked. Use when orchestrating codebase-heavy or token-heavy work with subagents, when dispatching Agent tasks without a model, when the user says "route to cheaper models", "efficient dispatch", "optimize model spend", or asks where their token spend goes. Pairs with the route-dispatch hook (deterministic backstop) and `ax dispatches` (evidence). Do NOT fire on single-shot questions or tiny tasks with no dispatching.
Necmttn/ax · ★ 107 · AI & Automation · score 78

Install: claude install-skill Necmttn/ax

# efficient-dispatch - routed, measured, verified The main model is the orchestrator and Q&A reviewer. Mechanical work runs on cheaper models - and unlike guidance-only approaches, every claim here is checkable against your own ax graph. ## The split Two axes. First, **main model vs subagent**: the main model orchestrates and reviews; mechanical work goes to subagents. Second, and the one that actually controls spend - **the tier of each subagent dispatch:** - **Implementer subagents** (well-specified plan tasks, mechanical edits, search, bulk transforms) → dispatch with **`model: sonnet`** (or haiku for pure search/locate, per the table). - **Reviewer / judgment subagents** (quality / PR / final / adversarial / code review, design, audit, architect, critique, judge) → **keep the strong model**: inherit the main model, or set `model: opus`/`fable` explicitly. Review is the catch-rate gate; a cheap reviewer misses real bugs. Get this backwards and you pay twice: in one ax session implementers ran on the expensive inherited model while reviewers were sent to a cheap one - ~$130 over, weaker catch rate, three fix rounds (memory `feedback-review-gets-strong-model`). The default-inherit trap is implementers, not reviewers: a forgotten `model:` on an `implement …` dispatch silently runs expensive. Set it. **Main model keeps** (never dispatched at all): decomposition, architecture and product tradeoffs, plan synthesis, judging conflicting subagent reports, final inte