claude-sdlc-kit

Reproducible parallel-orchestrator framework. Use when the user asks to "spawn an orchestrator", "run waves", "orchestrate <task>", "/orchestrate", "/integrate", or any multi-agent parallel build with a shared ticket queue. Loads PROTOCOL.md, LESSONS.md, and the matching template + agent profiles, then drafts a project-specific orchestrator prompt (does NOT spawn until the user approves). Do NOT use for single-task work or simple research delegation (use Agent directly).

Convert a multi-format BRD/FRD/spec corpus (.docx/.pptx/.pdf/.xlsx/.vsdx, often bilingual or scanned) into an agent-searchable markdown knowledge base. Use at the start of a "from BRD to tested app" build, when the requirements arrive as Office documents and you need a grep-able, provenance-tagged KB that downstream agents can cite. Also use when adding a new source document to an existing KB. Do NOT use for ordinary code or single-file conversions.

Use when designing or reviewing a database schema / data model - turning a knowledge base's data catalog (entities, fields, relationships) into a proper, normalized, constraint-backed schema before decomposing a spec into build tickets. Covers naming, data types, constraints, relationships, ORM/migration safety (expand/contract), indexing, transactions/concurrency, auditing, PII/compliance, JSON, partitioning, and anti-patterns. Syntax examples assume SQL Server + a TypeORM-style ORM; the principles are stack-agnostic.