Use when reviewing a gromit-next run output to decide whether to accept, reject and start fresh, or resume with targeted implementation guidance. Triggered by "review the run", "should I accept this", "what did the run produce", or any gromit-next run evaluation task.
Co-write a Gromit Next spec interactively. Explores the idea through structured questioning, then produces a well-scoped spec with vision, architecture, acceptance criteria, and behavioral scenarios. Works in Claude Code, Codex, or mobile apps.
Orchestrate the full Gromit pipeline from Claude Code. Shows pipeline dashboard, launches stages with fresh context, and dispatches work items. Use when working with Gromit projects.
Use when decomposing implementation plans into bd beads. Reads plans, extracts tasks, and outputs JSON array of bead definitions following strict sizing rules. Non-interactive, produces machine-readable output only.
Use when refining backlog items or rough ideas into structured specifications. Guides codebase-aware conversation to transform vague concepts into clear specs with acceptance criteria, decisions, and context.
Use when investigating bugs in the target codebase. Guides free-form investigation to identify root cause, triage severity, and produce appropriate outcomes (direct fix, plan, or backlog item).
Orchestrate the full Gromit pipeline from Claude Code. Shows pipeline dashboard, launches stages with fresh context, and dispatches work items. Use when working with Gromit projects.
Use when planning implementation for a specification. Reads specs, proposes architecture and test strategy with human review checkpoints, then writes a flexible implementation plan.