backlog
/backlog
Pilnuje formatu backlogu aplikacji: wycinki z wymaganiami, kryteriami zaliczenia i zależnościami oraz zasady wyboru kolejnego wycinka.
- Autor
- gtlpanda ↗
- Repozytorium
- PandaStack ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 12.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 2
- Skrypty
- 1
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 12.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
- 1 skrypt w paczce. Przejrzyj je przed instalacją, bo Claude może je uruchomić.
Skrypty: backlog.mjs
Co robi i kiedy
Pilnuje formatu backlogu aplikacji: wycinki z wymaganiami, kryteriami zaliczenia i zależnościami oraz zasady wyboru kolejnego wycinka.
Oryginalny opis (EN)
The format of an app's backlog (docs/backlog.json) - ordered /dev-sized slices with requirements, pass/fail acceptance criteria, and earlier-only dependencies - plus the build ledger and the rules for which slice runs next, when a slice is re-planned or blocked, and when an unattended build stops. Used by /blueprint (writes the backlog) and /build-app (builds it); use when writing, checking, or reasoning about a backlog.
Jak działa
- The format
- The ledger and the next step
Podgląd SKILL.md
---
name: backlog
description: The format of an app's backlog (docs/backlog.json) - ordered /dev-sized slices with requirements, pass/fail acceptance criteria, and earlier-only dependencies - plus the build ledger and the rules for which slice runs next, when a slice is re-planned or blocked, and when an unattended build stops. Used by /blueprint (writes the backlog) and /build-app (builds it); use when writing, checking, or reasoning about a backlog.
---
# backlog
A backlog turns an approved design into slices that `/dev` can build one at a
time, unattended. `/blueprint` writes it to `docs/backlog.json` in the app's
repo, after the human's one approval, and nothing else writes it. `/build-app`
reads it and keeps its progress in a ledger inside `.git`.
`backlog.mjs` in this skill owns every rule below. Check a file with:
```
node ~/.claude/skills/backlog/backlog.mjs check docs/backlog.json
```
It prints one line per problem, then `BACKLOG: PASS` or `BACKLOG: FAIL`.
`/dev --slice S<n>` gets its task from the same file:
```
node ~/.claude/skills/backlog/backlog.mjs task docs/backlog.json S2 [build state dir]
```
It prints `{slice, app, task, requiredCriteria}` as JSON: the task text (the
slice's goal and every criterion, plus each earlier reason and answer from the
build ledger) and the criteria the plan must carry unchanged. The ledger is
`<build state dir>/<app>/ledger.json`, one per app (`ledgerPath`).
## The format
```json
{
"version": 1,
"app": "reading-list",
"integrationBranch": "integration/reading-list",
"requirements": ["R1", "R2"],
"slices": [
{
"id": "S1",
"title": "Book list storage",
"goal": "Store books with title, author, and read state, and list them newest first.",
"requirements": ["R1"],
"acceptance": [
{ "id": "S1-AC1", "criterion": "A saved book is listed with its title and author", "fails_when": "the list is empty or shows another book after saving one" }
],
"depends_on": []
},
{
"id": "S2",
"title": "Mark a book read",
"goal": "Toggle a book between to-read and read, and keep it across restarts.",
"requirements": ["R2"],
"acceptance": [
{ "id": "S2-AC1", "criterion": "Marking a book read moves it to the read list", "fails_when": "the book stays in to-read after marking it" },
{ "id": "S2-AC2", "criterion": "Read state survives a restart", "fails_when": "a restart resets a read book to to-read" }
],
"depends_on": ["S1"]
}
]
}
```
Rules the check enforces:
- `app` is a lowercase slug, and `integrationBranch` is `integration/<app>`.
- `requirements` lists every requirement id of the design (`R<n>`), each once, and
every one is in at least one slice.
- Slice ids run `S1`, `S2`, ... in build order. A slice has a title, a goal, a
`requirements` list (it may be empty, for a scaffold slice), 1 to 8
acceptance criteria with ids `S<n>-AC1`, `S<n>-AC2`, ... and a non-empty
`fails_when` each, and `depends_on` naming only earlier slices, so there is
never a cycle.
- At most 40 slices.
- A goal never says "stop and ask", and neither does a title, criterion, or
`fails_when`: a slice is `/dev`'s task text, `/dev` lets task text override
its no-ask rule, and an unattended build has no one to ask.
What the check cannot see is the advisor's job at the backlog gate: whether
each slice is small enough for one `/dev` run, whether the criteria really
prove the requirements, and whether the order builds each owner before its users.
## The ledger and the next step
The `/build-app` driver keeps `.git/harness-build/<app>/ledger.json`, one
per app, never committed: per slice its status (`pending`, `done`, `blocked`), escalations,
re-plans, reasons, questions, and the test files it froze when it landed.
- **Next slice:** the first slice in order that is pending, not blocked, and
whose dependencies are all done. A slice that escalated since the last
landing goes after the others, so one stuck slice does not hold up
independent work.
- **Blocked:** a slice is blocked when only a human can unblock it (`/dev`
returned questions), or when it is stuck: it escalated twice, was re-planned
from scratch with every reason so far, and escalated twice again. Every slice
that depends on a blocked slice is blocked too. The rest keep building.
- **Frozen tests:** when a slice lands, its test files freeze. A later slice
that changes one does not land.
- **Brakes:** the build stops and pings the human on the first of
- a third re-plan in the app (the design or the backlog is wrong),
- three different slices escalating with none landing in between
(something shared is broken),
- every remaining slice blocked.
- **The way back in:** after a brake, `resume` starts the first two counters
over (the human has looked). A blocked slice stays blocked until `unblock`
gives it the human's answers; it then starts from a fresh plan with fresh
counts, and keeps its earlier reasons and every earlier answer.
Pliki w paczce (2)
- SKILL.md 5 KB
- backlog.mjs 18 KB
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · gtlpanda/PandaStack | 100 | MIT |
| Ovid/paad | 100 | MIT |
| Ovid/paad | 100 | MIT |
| Ovid/paad | 100 | MIT |
| SergioLizana/Dynamo | 100 | MIT |
| bitIO/wheelhouse | 100 | MIT |
| Archont561/qgis-agent-harness | 100 | MIT |
| umputun/cc-thingz | 100 | MIT |
| Alter-ai/skill-lab | 100 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki 24402a55a7f4 (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
- 90Backlog Management
Prowadzi backlog projektu w pliku BACKLOG.md: dodaje pozycje, aktualizuje je, sprawdza i ustala priorytety. - 100eventmodeling-planning-slices
Wypisuje wycinki dostawy modelu zdarzeń wraz z ich zależnościami, kontraktami i statusem, żeby zespoły wiedziały, co zbudować równolegle. - 90implementation-plan
Zamienia plan, PRD, element backlogu lub mgliste wymaganie w uporządkowaną pracę z kryteriami akceptacji. - 100user-stories
Tworzy priorytetową listę historyjek użytkownika z kryteriami akceptacji dla aplikacji na Windows, z briefu lub badań. - 100user-stories
Tworzy priorytetową listę historyjek użytkownika z kryteriami akceptacji dla aplikacji na macOS, z briefu lub badań. - 100user-stories
Tworzy priorytetową listę historyjek użytkownika z kryteriami akceptacji dla aplikacji na macOS, z briefu lub badań.