sdlc-plan
/sdlc-plan
Zamienia zatwierdzoną specyfikację w plan techniczny z punktami styku, ryzykiem, alternatywami, strategią testów i wdrożenia.
- Autor
- phongtn89 ↗
- Repozytorium
- sdlc-kit ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 13.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 13.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Zamienia zatwierdzoną specyfikację w plan techniczny z punktami styku, ryzykiem, alternatywami, strategią testów i wdrożenia.
Wyzwalacze z opisu
Argumenty
/sdlc-plan Path of spec.md (defaults to the active spec)
Oryginalny opis (EN)
Turn an approved spec into a technical plan (plan.md plus, only when needed, research.md, data-model.md, contracts/, ADRs): seams, blast radius, alternatives, failure modes, test strategy, rollout. Use after sdlc-clarify for T3 features, or when the user says 'plan this' or 'how should we build it'.
Jak działa
- When to use
- Inputs
- Process
- Rationalizations
- Red flags
- Verification
- Principles applied
Podgląd SKILL.md
--- name: sdlc-plan description: "Turn an approved spec into a technical plan (plan.md plus, only when needed, research.md, data-model.md, contracts/, ADRs): seams, blast radius, alternatives, failure modes, test strategy, rollout. Use after sdlc-clarify for T3 features, or when the user says 'plan this' or 'how should we build it'." argument-hint: "Path of spec.md (defaults to the active spec)" user-invocable: true --- # sdlc-plan: decide how, at the real seams ## When to use - T3, once `spec.md` has no blocking markers. - When the architecture of an existing feature changes. Update its `plan.md` and add an ADR. - Not for T0–T2. Their design fits in the mini-spec or the PR description. ## Inputs - `spec.md`, `.sdlc/constitution.md`, `.sdlc/config.json`, `AGENTS.md`, ADRs, glossary. - `.sdlc/templates/plan.md` and `.sdlc/templates/adr.md`. - Checklists, loaded only when triggered: `.sdlc/checklists/security.md`, `.sdlc/checklists/observability.md`, `.sdlc/checklists/performance.md`, `.sdlc/checklists/compatibility-migration.md`, `.sdlc/checklists/accessibility.md`. ## Process 1. **Read the seam, not the repo.** For each scenario, find the entry point and trace its callers and consumers: graph tools if available, otherwise bounded search. Delegate wide sweeps to `sdlc-explorer`. Record the blast radius: modules, contracts (APIs, events, schemas, SDKs), migrations, other repos, with `path:line`. 2. **Inventory what exists.** List the types, schemas, helpers and dependencies the solution can reuse. Read `.sdlc/principles/subtract-before-adding.md`. Canonical contracts owned elsewhere are imported, never re-declared. 3. **Compare approaches.** Write a table of 2–3 credible options: how each meets the FRs, risk, cost, compatibility. Pick one and say why. Put a hard-to-reverse choice in an ADR (`docs/adr/NNNN-*.md`, status `proposed`). 4. **Fill `plan.md`** from the template, keeping each section only if it earns its place: - Technical context: runtime, dependencies, storage, test tooling, constraints. Unknowns become `research.md` items, resolved before tasks. - Constitution check: every principle passes, or the violation is justified under *Complexity justifications*. - Failure modes: timeouts, retries, idempotency, concurrency, partial failure, replay, ordering, rollback. Cover only the cases that can actually occur. - Security and observability: run the triggered checklists and note the outcomes. - Test strategy: the test seam for each scenario (public interface), the RED command for each, and the integration or e2e boundary where mocks can't prove the behaviour. - Rollout and rollback: order of deploy across services, flags, the migration expand/contract steps. 5. **Contracts and data, only if the feature changes them.** Write `data-model.md` (entities, ownership, invariants, state transitions) and `contracts/` (request, response and event shapes, error cases, versioning). If the contract lives in another repository, the plan names that dependency and its owner. Don't fork it locally. 6. **Size it.** Estimate the changed lines and the seams against `.sdlc/config.json` → `budget`. Above `reassess_lines`, simplify first. Split only into deliverables that are independent (separate owner, repo or risk), never into "code now, tests later". 7. **Report**: the plan path, the chosen approach in 2 lines, any ADRs drafted, open research items, size estimate, and the next step `sdlc-tasks`. ## Rationalizations | Excuse | Reality | |---|---| | "I know the codebase, skip tracing callers" | Most regressions come from consumers nobody traced. Trace them and cite them. | | "One option is obviously right" | Then the comparison takes two lines. Writing it down catches the times it isn't. | | "Add an abstraction for future needs" | Speculative extension points are complexity you pay for now. Use the decision ladder. | | "Define our own copy of the DTO" | Parallel contract types drift. Import the canonical schema, or record the missing dependency. | | "Failure modes are an ops problem" | Retries, idempotency and partial failure are design decisions. Make them here or debug them in prod. | ## Red flags - The plan needs a change in another repository's contract. Record the dependency and its owner. Consider T4. - Research items are still open but tasks are about to be written. Resolve them first. - The constitution check fails without a justification. - The size estimate is far above budget and no simplification has been tried. ## Verification - Every FR in the spec maps to a part of the plan. - The blast-radius list cites real paths, and consumers were checked. - The test strategy names a seam and a RED command for each P1 scenario. - The triggered checklists were run, with outcomes recorded. - No open research items remain, or they are listed as blocking with an owner. ## Principles applied subtract-before-adding, depth-follows-risk, evidence-over-claims, guard-context
Wersje
W bazie nie ma innych skilli o nazwie sdlc-plan.
Migawka bazy z 13.11.2026 · wersja paczki f6a19fb2bd69 (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
- 100sdlc-architecture-decision-record
Zapisuje ADR-y z wyborami, kompromisami, alternatywami, konsekwencjami i statusem, z linkami do wymagań i dowodów projektowych. - 100plan
Zamienia zatwierdzoną, doprecyzowaną specyfikację w techniczny plan przed kodem: architektura, kontrakty, przepływ danych, strategia testów, kolejność prac i ryzyka na poziomie struktury. - 90pre-flight
Uruchamia nowy projekt programistyczny w metodzie AI Native SDLC, zamieniając intencję w weryfikowalną specyfikację, plan i dowody, z regułami w AGENT.md. - 90sdlc-specify
Tworzy lub aktualizuje specyfikację funkcji, czyli co i po co, bez implementacji, w pliku spec.md w katalogu specs. - 90sdlc-clarify
Zadaje najwyżej pięć pytań, żeby wyjaśnić otwarte kwestie w szkicu specyfikacji, i zapisuje odpowiedzi z powrotem do dokumentu. - 100sdlc-design
Tworzy dokument projektu technicznego z alternatywami rozwiązań, modelem danych, kontraktami interfejsów i tabelą decyzji.