shape
/shape
Kształtuje zapis zadania przed budową: zamiar prostymi słowami, numerowane wymagania i listę dowodów w pliku SPEC.md, a także tryb przesłuchania.
- Autor
- vanducng ↗
- Repozytorium
- skills ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 13.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 22
- Skrypty
- 6
- 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.
- 6 skryptów w paczce. Przejrzyj je przed instalacją, bo Claude może je uruchomić.
Skrypty: evals/01-thin-record/checks.sh, evals/02-options/checks.sh, evals/03-grill/checks.sh, evals/04-ready-record/checks.sh, evals/05-stateful-brownfield/checks.sh, evals/05-stateful-brownfield/fixture/src/upload.js
Co robi i kiedy
Kształtuje zapis zadania przed budową: zamiar prostymi słowami, numerowane wymagania i listę dowodów w pliku SPEC.md, a także tryb przesłuchania.
Wyzwalacze z opisu
Argumenty
/shape [--grill | --options] [record dir]
Oryginalny opis (EN)
Shapes a work item's record before it is built: the intent in plain words, numbered mini-requirements anyone can check, and a checkbox proof contract, in the record's SPEC.md. A thin router: --grill walks an existing plan one question at a time; --options lays out two or three approaches and picks one. Use when asked to 'shape this', 'spec this', 'write the requirements', 'what is the proof contract', 'grill me', 'stress-test this plan', 'what are our options', 'how should we approach this', or 'plan the build'. Long multi-session decision maps stay with vd:interview --wayfinder.
Jak działa
- Modes
- Default steps
- Rules
Podgląd SKILL.md
---
name: shape
description: "Shapes a work item's record before it is built: the intent in plain words, numbered mini-requirements anyone can check, and a checkbox proof contract, in the record's SPEC.md. A thin router: --grill walks an existing plan one question at a time; --options lays out two or three approaches and picks one. Use when asked to 'shape this', 'spec this', 'write the requirements', 'what is the proof contract', 'grill me', 'stress-test this plan', 'what are our options', 'how should we approach this', or 'plan the build'. Long multi-session decision maps stay with vd:interview --wayfinder."
license: MIT
argument-hint: "[--grill | --options] [record dir]"
metadata:
author: vanducng
version: "0.2.0"
---
# shape
The second gate. A record leaves it with an intent, mini-requirements and a proof contract a reviewer can check without asking you. It does not build.
**The record:** in an OpenRig rig, the slice folder. Elsewhere, the bench record the workbench skill resolves (`records.cjs resolve`).
## Modes
| Mode | When | Reads | Writes |
|---|---|---|---|
| default | the record exists, and its requirements or contract are thin or missing | `references/sequencing.md` when the build takes more than one sitting | `SPEC.md`, plus `PROGRESS.md` `## Plan` if sequenced |
| `--grill` | a plan or idea exists and has to be stress-tested | `references/grill.md` | nothing until the answers are in, then `SPEC.md` |
| `--options` | the approach is not obvious | `references/options.md` | `reports/options.md` |
## Default steps
1. **Read the record.** If `SPEC.md` already meets steps 2 to 6, change nothing: say it is ready and name the next gate (build, then `vd:code-review` and `vd:verify`).
2. **Intent:** one or two sentences in plain words: who gets what, and why.
3. **Mini-requirements:** numbered, observable, testable outcomes ("`total()` counts a repeated coupon once"), not tasks ("refactor cart"). Each one maps to a contract line, its success check.
4. **Proof contract:**
- keep what intake wrote; none there, build the lines from the single source: `node <intake>/scripts/contract.cjs <type>`;
- make every line concrete: name the test, the surface or the number;
- add the door type (one-way: hard to undo, such as data or a public API; two-way: a revert undoes it), the blast radius and the undo step if missing.
5. **States,** only when the frontmatter names the `statechart` lens. With `vd:statechart`:
- for work on existing code, two charts: **today**, modeled from the code, in `reports/statechart/today/`, and **intended**, from the rules, in `reports/statechart/intended/`; new work gets the intended chart only;
- each rule is a numbered mini-requirement ending `(invariant)` and an `INVARIANT` of the intended model. A rule the ask states is approved; ask the reader before adding any other;
- one contract line per invariant and per bug path the today chart marks, each a test that fails on today's code and passes after the change;
- the charts' paths in `SPEC.md`, under `## States`.
6. **Out of scope:** one line, so the reviewer knows what not to ask for.
7. **Frontmatter, when intake left none:** `id`, `type`, `intent`, `depends_on` (the slices this waits on).
8. **Territories:** when the build spans repos, one worktree per repo and one PR by default. Name them here.
9. **Check:** no `TODO` is left. The record must pass `rig scope audit` in a rig, or the workbench `check` elsewhere.
## Rules
- **Never invent a requirement the ask does not support.** If one is needed, ask one question and give your best guess with it.
- **One question at a time,** in every mode.
- **Shape writes the record only.** Code, plans in new files and prototypes are out of scope; the statechart model and charts under `reports/statechart/` are part of the record.
Pliki w paczce (22)
- SKILL.md 4 KB
- evals/01-thin-record/checks.sh 1 KB
- evals/01-thin-record/prompt.md 1 KB
- evals/01-thin-record/fixture/.workbench/cart-coupon/SPEC.md 1 KB
- evals/02-options/checks.sh 1 KB
- evals/02-options/prompt.md 1 KB
- evals/02-options/fixture/.workbench/price-cache/SPEC.md 1 KB
- evals/03-grill/checks.sh 1 KB
- evals/03-grill/prompt.md 1 KB
- evals/03-grill/fixture/.workbench/price-cache/SPEC.md 1 KB
- evals/04-ready-record/checks.sh 1 KB
- evals/04-ready-record/prompt.md 1 KB
- evals/04-ready-record/scenario.json 1 KB
- evals/04-ready-record/fixture/.workbench/cart-coupon/SPEC.md 1 KB
- evals/05-stateful-brownfield/checks.sh 1 KB
- evals/05-stateful-brownfield/prompt.md 1 KB
- evals/05-stateful-brownfield/scenario.json 1 KB
- evals/05-stateful-brownfield/fixture/.workbench/upload-cancel/SPEC.md 1 KB
- evals/05-stateful-brownfield/fixture/src/upload.js 1 KB
- references/grill.md 1 KB
- references/options.md 1 KB
- references/sequencing.md 1 KB
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · vanducng/skills | 90 | MIT |
| biliboss/skills | 100 | MIT |
| gmdkaio/kstack | 100 | MIT |
| ebullient/qrspi-x | 100 | Apache-2.0 |
| YnotnaUK/ai-skills | 100 | MPL-2.0 |
| jamesm-dev/payday-dev | 100 | MIT |
| nyelonong/galdr | 100 | MIT |
| thecraighewitt/skills | 100 | MIT |
| tiago-peixoto/claude-shapeup | 95 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki bb83b799f5f8 (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
- 100grill
Łączy dwa tryby: stress-test planu lub projektu aż do ujawnienia luk oraz shape, czyli iteracyjne dopracowanie wymagań i opcji rozwiązania do uzgodnienia. - 90intake
Zamienia nowe zlecenie w pozycję pracy: nazywa jej typ, streszcza zamiar i zapisuje SPEC.md z dowodami akceptacji. - 90grill
Dociekliwie przepytuje użytkownika o plan lub decyzję, zanim zacznie się wdrażanie. - 90grill-me
Przepytuje plan, decyzję lub pomysł bez litości, by sprawdzić myślenie przed budową. - 95speckit-grill-me
Przepytuje użytkownika ze specyfikacji funkcji spec-kit, aż do wspólnego zrozumienia każdej gałęzi decyzji. - 100grill-me
Przepytuje użytkownika w numerowanych rundach z rekomendacją, aż żadna decyzja nie zostanie przyjęta po cichu.