just-claude-things

Use as the last planning gate before scaffolding — stress-tests a completed plan (requirements + tech-design + tickets) for failure modes via a blind dual-draft pre-mortem, then emits a failure register with mitigations or accepted risks. Requires complete tech-design and issues docs. Triggers include "pre-mortem", "premortem", "how could this fail", "what could go wrong", "stress-test the plan", "risk the plan before we build".

Use when stress-testing or eliciting pure functional requirements for a product or feature. Drives the *what*, not the *how*. Refuses to discuss tech stack, architecture, or non-functional requirements. Triggers include "grill requirements", "what should this do", "define the what", "elicit requirements", "PRD interview".

Use when starting a new project after tech design is complete. Generates the repo scaffold (optionally populated from tech-design XML tags — Phase-2), root CLAUDE.md, per-module AGENTS.md with IDed architectural rules, matching semgrep rules, NetArchTest arch tests, and Claude-only git hooks. Triggers include "scaffold the repo", "set up the project", "wire in arch tests", "bootstrap new app".

Use when designing a system after requirements are settled. Drives architecture, components, data model, API contracts, NFRs, and ADRs out of the room. Requires a complete requirements doc. Triggers include "tech design", "design the system", "how should we build this", "architecture session".

Use when ready to break complete requirements + tech design into independently-shippable, tracker-agnostic issues. Outputs a markdown file using tracer-bullet vertical slices. Triggers include "break into tickets", "create issues", "slice the work", "issue list", "what should we build first".