nauro-looplisted

# Nauro loop skill Run a gated iteration of work origination on top of `/nauro-ship-task`. This skill is a thin outer loop. It mines the project's existing Nauro store state for candidate work, ranks a small set, and surfaces it to the human to pick; on the human's pick it dispatches `/nauro-ship-task <chosen task>` byte-for-byte with all six inner gates intact, then loops back. It originates the candidate set; the human selects, approves, and confirms everything downstream. The loop adds one net-new agent authority — task origination — and nothing else. Today the human authors the task description handed to `/nauro-ship-task`; the loop now proposes and ranks the candidate "what to build next" set. It enumerates options; it never selects. Everything past enumeration stays with the human: the SELECT ratify-gate, the chain's plan-approval gate, every AMBER/RED tech-lead pause, and every push confirmation. ## What the loop cannot do These are structural, not stylistic. The loop holds none of these capabilities and must not simulate them. - The loop never files a decision. It holds no doctrine-write authority into the store, cannot record doctrine, and carries no path to commit one. It runs in the main-agent context with no tool-lock, so an autonomous filing would install binding doctrine with no human gate; the only filing that ever happens is inside the dispatched chain, after a human clears a chain gate, by the agent the chain assigns. Writing a SELECT checkpoint (below)