atlas-verify
/atlas-verify
Sprawdza, czy praca jest naprawdę skończona: uruchamia build, sprawdzanie typów, linter i testy, a każdy wynik raportuje osobno.
- Autor
- Z0CC0 ↗
- Repozytorium
- ATLAS-max ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 03.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 6
- Skrypty
- 1
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 03.11.2026, 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: workflows/ship.mjs
Co robi i kiedy
Sprawdza, czy praca jest naprawdę skończona: uruchamia build, sprawdzanie typów, linter i testy, a każdy wynik raportuje osobno.
Wyzwalacze z opisu
Oryginalny opis (EN)
Says whether the work is actually done: runs the project's build, type check, linter and tests, looks for what was left behind, and reports each as passed, failed, could not run or not run. Angles by the request: "before release", "production", "launch" (readiness audit); "buttons", "click path", "UI state" (handlers that undo each other); "ship check", "run the checks in parallel", "use a workflow" (every check by its own runner beside a security review, one verdict). Use for "atlas verify", "is it ready", "did it work", "check everything before I commit" — in any language.
Jak działa
- What to read
- Who runs the commands
- Ship check — a workflow
- Boundaries
- Form
Podgląd SKILL.md
---
name: atlas-verify
description: >
Says whether the work is actually done: runs the project's build, type check, linter and
tests, looks for what was left behind, and reports each as passed, failed, could not run or
not run. Angles by the request: "before release", "production", "launch" (readiness audit);
"buttons", "click path", "UI state" (handlers that undo each other); "ship check", "run the
checks in parallel", "use a workflow" (every check by its own runner beside a security
review, one verdict). Use for "atlas verify", "is it ready", "did it work", "check
everything before I commit" — in any language.
---
"Done" is a claim. This skill replaces it with what was run and what came back.
## What to read
Everything below lives in the `verify/` folder beside this file.
1. `verify/method.md`, always: the chain of checks, the four words, the comparison with the
last run, the report.
2. By the words of the request, on top of it:
"release", "production", "launch", "deploy", "go live", "ship" → `release.md`
"buttons", "click", "handlers", "UI state", "nothing happens when I press" → `clicks.md`
"coverage", "threshold", "the bar", "ratchet", "exception", or a project with thresholds
in CI → `bar.md`
A deployed URL → that is `atlas-test` in its site mode; say so and hand over.
## Who runs the commands
The `atlas-runner` subagent, when it exists, one command per call: the verdict and the
deciding lines come back, the logs do not. Without it: output to files in the scratch
directory, read the deciding lines only.
## Ship check — a workflow
"ship check", "run the checks in parallel", "use a workflow", "fan out the checks": the
multi-agent form, and that request is the opt-in the Workflow tool requires. First the
work-list, scouted here: the project's own check commands from `package.json` scripts,
`Makefile`, `pyproject.toml`, `Cargo.toml`, `go.mod` or the CI file (test, type check, lint,
build), quoted as the project writes them — none found, say so, nothing runs on nothing; the
diff target (`git diff`, `git diff --staged`, `git diff <base>...`). Say the cost in one line:
one runner per command plus one reviewer. Then one call, with the plugin prefix (the name
before the colon in the command used) so the script reaches `atlas-runner` and `atlas-diff`:
```
Workflow({ scriptPath: "${CLAUDE_PLUGIN_ROOT}/skills/atlas-verify/workflows/ship.mjs",
args: { commands, target, prefix } })
```
It returns `lines`: one `PASS` / `FAIL` / `DID NOT RUN` per check, the security lines with
their verdict, and `ship` or `do not ship` with the reason; print them unchanged. A check
that returned nothing is `DID NOT RUN`, never assumed green. Without the Workflow tool: the
chain in `method.md`, one command at a time.
## Boundaries
Verifying changes nothing. A failed check is reported, not repaired: `atlas-fix` repairs a
build, `atlas-review` finds what is wrong in a diff, the user decides. No file is edited, no
dependency installed, nothing committed.
## Form
The active compression level governs the prose. It never overrides the report format in
`method.md`. Error strings are quoted exactly.
Pliki w paczce (6)
- SKILL.md 3 KB
- verify/bar.md 2 KB
- verify/clicks.md 2 KB
- verify/method.md 3 KB
- verify/release.md 3 KB
- workflows/ship.mjs 5 KB
Wersje
W bazie nie ma innych skilli o nazwie atlas-verify.
Migawka bazy z 13.11.2026 · wersja paczki bbcd6aaa1bed (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
- 85atlas-fix
Doprowadza do zielonego stanu błędną kompilację, sprawdzanie typów lub linter najmniejszymi zmianami, po jednym błędzie naraz. - 100check
Uruchamia tylko lint, sprawdzanie typów, build i testy związane ze zmianą i raportuje prawdziwe wyniki. - 100atlas-review
Przegląda kod pod kątem błędów i ryzyk w diffie, gałęzi lub pull requeście, z poprawkami i listą problemów. - 100atlas-review
Wskazuje w kodzie, co się zepsuje i przy jakim wejściu, wraz z poprawką, pomijając elementy poprawne. - 100atlas-review
Wypisuje problemy w diffie, gałęzi, pliku lub pull requeście, po jednym w wierszu, z miejscem i poprawką. - 100ship
Sprawdza, czy praca jest naprawdę skończona, potem zapisuje zmiany w commicie, otwiera PR, kończy gałąź i wdraża zmianę.