blueprint-vs-rawdogging

Use when writing or updating an ADR for a specific technical decision.

Use when creating, reviewing, or restructuring AGENTS.md, CLAUDE.md, or scoped instruction files for a project, including root, feature, service, backend, or domain-level guidance.

Use when a new app plan already exists and needs an interactive kickoff into setup decisions and first tasks.

Use at task handoff, milestone review, or after implementation when project lessons may be reusable for future apps, toolkit docs, AGENTS/CLAUDE instructions, skills, or templates.

Use when a greenfield app needs a mature foundation plan before implementation.

Use when explicitly defining MVP scope, exclusions, beta screens, states, or success metrics.

Use when drafting pull request descriptions, end-of-session handoffs, reviewer notes, verification summaries, or stop-start task updates.

Use when centralizing runtime config, environment separation, or fail-fast env validation.

Use when an uncertain external system needs a typed service boundary and realistic stub.

Use when explicitly choosing a stack or documenting a deviation from toolkit defaults.

Use when reviewing AGENTS.md, CLAUDE.md, skills, prompts, docs, or repo workflow for token bloat, context inefficiency, broad scans, overlong instructions, or unnecessary subagent usage.