system-design-plan
/system-design-plan
Przygotowuje szczegółowy plan implementacji i dokument architektury do przekazania innemu inżynierowi lub agentowi.
- Autor
- JanPer1389 ↗
- Repozytorium
- system-design-skill ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 08.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 08.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Przygotowuje szczegółowy plan implementacji i dokument architektury do przekazania innemu inżynierowi lub agentowi.
Oryginalny opis (EN)
Use when the user requests a detailed implementation plan or architecture handoff for another engineer or agent.
Jak działa
- Planning sequence
- Detail and completion
- Technical judgment and presentation
Podgląd SKILL.md
--- name: system-design-plan description: Use when the user requests a detailed implementation plan or architecture handoff for another engineer or agent. metadata: version: "0.4.0" --- # Write a system implementation plan **Core principle:** transfer decisions an engineer cannot infer: exact responsibilities, interfaces, values and checks. A plan is neither a vague task list nor the full application written in advance. Read [shared policies](../system-design-enter/references/policies.md), [operating method](../system-design-enter/references/methodology.md) and **[planning method](../system-design-enter/references/planning-method.md)**. Use [the worked excerpt](../system-design-enter/references/planning-example.md) to calibrate specificity. Conceptual name: `system-design:plan`; portable ID: `system-design-plan`. ## Planning sequence 1. **Establish sources and the review gate.** Missing architectural decisions **OR** missing code-grounded understanding requires `../system-design-review/SKILL.md`. Discussion alone is insufficient. Without code review requirements/design once, state implementation absence and continue. Reuse current sufficient evidence; do not loop. Brainstorm requires consolidated-design confirmation before this handoff. 2. **Establish the system map and target.** Map goals, actors, domains, ownership, workflows and external dependencies. Independently test the supplied architecture against requirements; distinguish binding constraints from suggestions/current code. Apply stack assessment, compare the smallest sufficient alternatives and choose a justified target, including structural improvements where needed. Explain cost and behavior changes in user terms. Add diagrams under the shared triggers. Then **map scope and file responsibilities.** Inspect stack, conventions and exact symbols as text. Group changes by responsibility and coherent deliverables; preserve justified architecture. State each file's ownership before task decomposition. Completion: file structure has reasons and public boundaries. 3. **Define global constraints and contracts.** Link design/review; specify goal/status, architecture, stack, binding values, assumptions, types, errors, state owners and dependencies. Establish transaction/repeat/concurrency semantics. Add review-focus failure/input classes with owning tasks/checks. 4. **Size independently verifiable tasks.** Each includes result/requirement, exact Create/Modify/Test paths, interfaces consumed/produced, named dependencies and checkbox steps with checkable outcomes. Assign each helper/fixture/type to existing code or a task. Identify independent tasks and shared-file conflicts without automatically assigning agents/worktrees/commits. 5. **Write behavior steps in TDD order.** Write concrete test assertions ? run the exact command and confirm expected failure ? implement the minimum specified interface/algorithm ? focused/regression checks ? useful refactoring and recheck. State working directory and expected observation. Record an explicit user override; documentation gets meaningful validation rather than fake failing tests. 6. **Carry evolution through tasks.** Address migration, compatibility, coexistence and recovery where applicable. Put concurrency, duplicate delivery and timeout checks in the task owning the invariant, not generic advice. 7. **Self-review against sources.** Trace requirements to tasks/checks; scan vague steps and excessive bodies; compare types/names/values; check review-focus coverage, consistency semantics, order and file conflicts. Correct supported gaps before handoff. ## Detail and completion Use [domain discovery](../system-design-enter/references/domain-discovery.md) for unclear responsibilities; [relationships](../system-design-enter/references/context-relationships.md) and [mechanisms/frameworks](../system-design-enter/references/relationship-mechanisms.md) for contracts crossing owners; [pattern selection](../system-design-enter/references/pattern-selection.md) for choices. Activate `../system-design-microservices/SKILL.md` for service concerns, reusing it if active. Show test assertions, exact signatures and essential algorithm excerpts; leave routine bodies to the executor. Red flags: undefined types, invented line numbers, vague testing instructions, commands without expected results or a program transcript disguised as a plan. Present the substantive plan in the host plan view when available, otherwise in chat: system map, target architecture/stack and rationale, exact tasks/interfaces/checks, acceptance criteria, assumptions, risks and unknowns. A link or short summary alone is not the handoff. Save a matching copy under the shared plan path when permitted; report unsaved status if writing fails. Follow the shared output-limit and user-preference rules. Commands are **unexecuted handoff instructions**. Do not run them, start implementation, change application files/dependencies/infrastructure or ask a redundant execution-method question. Outside brainstorm ask only about a blocking material decision. ## Technical judgment and presentation Apply the shared agent-owned decision and diagram policies. Use [stack assessment](../system-design-enter/references/stack-assessment.md) for material architecture/tooling choices and [diagram guidance](../system-design-enter/references/diagrams.md) when a drawing is requested, substantial work benefits from one, or a prompt-only model has enough facts. Choose technical details yourself; explain product consequences without asking the user to select technologies. Preserve binding constraints, not unexamined suggestions. Keep this operation's scope and nonimplementation limits.
Wersje
W bazie nie ma innych skilli o nazwie system-design-plan.
Migawka bazy z 13.11.2026 · wersja paczki 3a338fcec8d4 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100plan
Tworzy szczegółowy plan implementacji funkcji lub zadania, gdy użytkownik chce planować przed kodowaniem. - 90plan
Przygotowuje plan architektury funkcji przed implementacją, gdy są już kryteria akceptacji i specyfikacja. - 100writing-plans
Tworzy szczegółowy plan wdrożenia na podstawie specyfikacji, dokumentu projektowego lub wymagań, z kolejnością implementacji, przed rozpoczęciem kodowania. - 100make-plan
Tworzy szczegółowy fazowy plan wdrożenia z odkrywaniem dokumentacji, zwłaszcza przed wykonaniem poleceniem do, dla funkcji, zadania lub wieloetapowej implementacji. - 100create-implementation-plan
Tworzy plik planu wdrożenia dla nowych funkcji, refaktoryzacji, aktualizacji pakietów lub zmian architektury, z weryfikacją każdej fazy i podsumowaniami przekazania dla kolejnych agentów. - 100plan-crafting
Tworzy szczegółowy, wieloetapowy plan wdrożenia z zatwierdzonego projektu lub ustalonych wymagań, zanim zaczną się zmiany w kodzie.