Plan, build, review, and verify the NEW greenfield TOPOz V1 from its PRD. Use for architecture, task decomposition, implementation, QA, scrape/crawl/search/documents/OCR/jobs/API/UI/policies, and optional AI Mode. Enforces no modification/reuse of old TOPOz code, Python-first deterministic processing, explicit AI opt-in, strict contracts, phase gates, and automated evidence.
Resources
2Install
npx skillscat add tresdtres/topoz/topoz-greenfield-builder Install via the SkillsCat registry.
TOPOz Greenfield Builder
Read first
AGENTS.md, docs/PROJECT-STATE.md, docs/PRD-V1-EXTRAS.md, docs/ARCHITECTURE-CONTRACT.md, docs/GREENFIELD-RULES.md, docs/QA-GATES.md, docs/DECISIONS.md.
Absolute rule
This is a new implementation. Do not modify, migrate, refactor or automatically copy previous TOPOz source.
PLANNING
Do not create product implementation. Generate/refine docs/IMPLEMENTATION-PLAN.md, project-tasks/TASKLIST.md, architecture/contract/test-plan docs, and OPEN decisions. Planning must be dependency ordered, contract-driven and testable. Do not change the phase yourself.
IMPLEMENTATION_APPROVED
For each approved task: bind PRD/task ID; classify DETERMINISTIC or AI_OPT_IN; state input/output/error contracts; implement only the bounded task; add tests; run python .agents/skills/topoz-greenfield-builder/scripts/verify_project.py plus repository checks; record evidence; do not silently start the next task.
Deterministic work: no LLM SDK/provider call, fallback, credential requirement or provider network dependency.
AI work: provider calls only inside AI boundary; consume processed/base content; validate structured output; record non-secret usage metadata.
Review priority: PRD violations > hidden AI coupling > contract drift > security boundary > missing tests > style.
Prohibited: old-code migration/copying without explicit authorization; LLM fetch/crawl/HTML cleanup/OCR/link discovery/ordinary parsing; scope expansion; silent schema changes; completion claims without evidence.