Software Slice Selection

Overview

Mark the just-completed slice or slice batch (if any), update the completed slices list, and select the next unfinished execution unit from a software delivery plan stored in .stencila/plans/. Given a just-finished slice name and a list of already-completed slices, append the finished slice or slices, identify the next unit of work to execute based on phase ordering and dependency constraints, validate its package references against the codebase, and report the unit details along with the updated completed list. If all slices are complete, report that no slices remain.

This skill determines slice exhaustion, whether there are more execution slices to select, not full plan completion. Many plans include a ## Definition of Done section with plan-level closure checks (e.g., all tests passing, linting clean, generated types updated, documentation complete). This skill does not evaluate those checks. When all slices have been selected and completed, the skill reports that no slices remain, but the caller must verify Definition of Done items separately before treating the plan as finished.

This skill reads a plan file, parses its structure, tracks completions, and determines what comes next. It does not write code, create tests, modify the plan, or manage any external state beyond reporting its findings. It also acts as a normalizer of planning granularity: plan-authored slices are advisory, and the selected execution unit may consist of one slice or several adjacent compatible slices when that would reduce workflow overhead without creating an unmanageably broad TDD loop.

Required Inputs