spec-kitty-runtime-next

What problem does it solve?

This Skill helps you reliably advance a Spec Kitty mission step-by-step, including handling WP-level iteration and doctrine-scoped context, so you can keep progress moving and interpret runtime decisions correctly.

Core Features & Use Cases

• Canonical runtime control loop: Drives the mission forward using the single entry point spec-kitty next and the returned decision kind (step, decision_required, blocked, terminal).
• WP iteration without step advancement: Ensures implement/review steps can return different WP IDs while keeping the same step_id until all WPs reach terminal/handoff lanes.
• Doctrine-aware, action-scoped context loading: Loads agent profile at init and refreshes governance context at each step boundary, with bootstrap-vs-compact loading and on-demand directive/tactic pulls.
• Decision interpretation and safety checks: Guides handling of guard_failures, prompt_file null cases, and the “completed missions return step” workaround.

Quick Start

Use the AI to run the canonical loop by repeatedly calling spec-kitty next --agent <agent> --mission <mission-slug> --json, executing the returned prompt when kind is step, answering when kind is decision_required, resolving blockers when kind is blocked, and running final acceptance when kind is terminal.