What problem does it solve? Implementing a multi-step Objective by hand requires repeatedly deciding what slice to build next, dispatching work, verifying results, and keeping commits and tracking honest. This Skill automates that loop: it selects an Objective, previews the run, dispatches one implementation step at a time, and judges each result against real repository evidence before accepting it. ## Core Features & Use Cases - Dual execution modes: ns-bookended mode uses runner-begin/runner-finish for runner-attested checkpoints, while portable mode works with plain Git and parent-verified ordinary commits. - Parent-judged step loop: each step gets thin guidance, exactly one implementation dispatch, and evidence-based accept/recover/stop judgment, with a hard step-budget ceiling. - Honest run reporting: finishes with a structured Autorun digest separating runner-attested facts, parent-verified evidence, child claims, tracking commits, and publication outcomes. - Use Case: You have an Objective recorded under .ns/objectives/ with a roadmap. Ask the agent to autorun it for 3 steps; it previews the branch topology and verification authority, implements each slice, verifies the diff itself, commits accepted work on a dedicated run branch, and ends with a digest of what changed. ## Quick Start Ask the agent to autorun the Objective at .ns/objectives/my-feature for up to three steps and confirm the launch preview it shows you.
