← ClaudeAtlas

oc-configuring-model-routinglisted

Interactive skill to reconfigure harness.routing.json via dual-safe presets or custom slots. Deterministic apply rewrites routing + all agent frontmatter models + AGENTS.md §8 + opencode.json model/small_model. Validates before any write. Never disables dual without explicit operator override. Never invents roles or commits secrets.
orobsonn/claude-harness · ★ 1 · AI & Automation · score 73
Install: claude install-skill orobsonn/claude-harness
# Configuring-Model-Routing **This skill reconfigures model routing. It does not implement features.** Runs interactively inside the root `build` conversation. The operator may type `/configuring-model-routing` or make a direct request — operator messages in **pt-br product-language**; file content in English. Announce at start (pt-br): "Vamos ajustar quais modelos cada papel do harness usa." **No ceremony.** Reconfiguring routing is a harness-lifecycle op, not a product delivery. It runs before triage in root `build` (no `classify`, no `oc-brainstorming`, no planner/adversary); the native tool and entry gate are the safety net. **Apply via the native tool — never `node -e`, never hand-edit the touchpoints.** The `configure-routing` tool wraps the sanctioned engine in-process with validation, staged-write, and rollback across every touchpoint. Hand-editing routing with `sed`/`perl` skips all of that and risks leaving touchpoints out of sync — bash can technically reach the files, but use the tool anyway. - `configure-routing({ action: "inspect" })` → presets + touchpoints + current routing (read-only). Use for step 1. - `configure-routing({ action: "apply", preset })` or `({ action: "apply", slots })` → validate + staged-write + rollback across all touchpoints. Engine internals live in `skills/configuring-model-routing/references/apply-routing.mjs` (`listPresets`, `buildRoutingFromSlots`, `routingFromPreset`, `applyRoutingToDisk`, `listRoutingTouchpoints`) — the tool