What problem does it solve? Feature planning often produces vague specs that executors cannot act on without the original conversation, and acceptance criteria get written after implementation instead of before. This Skill runs planner sessions that translate a human's raw intent into a signed intent spec, one work brief per milestone, and runnable acceptance tests authored before any code exists. ## Core Features & Use Cases - Plan mode: Interview the human, investigate the codebase, decompose the feature into vertical-slice milestones, pin each brief's Surface, and author acceptance tests measured against the real runtime before sign-off. - Triage and replan modes: Verify executor divergence reports against the code, classify them as false facts, untenable decisions, or wrong outcomes, and re-draft affected briefs and tests with an amendment log. - Close mode: Run a fresh-session intent-conformance review over the whole feature diff, produce a close report, and archive the spec with acceptance tests promoted by kind. - Use Case: A developer types an intent dump for a new reminders feature; the planner walks them through the relevant code, asks the unresolved questions, writes SPEC.md plus per-milestone briefs and failing acceptance tests, gets per-item sign-off, and hands off to the executor workflow. ## Quick Start Ask the AI to plan a new feature by describing your idea, for example: run kspec to plan adding wellness reminders to the app and walk me through the decisions.