setup
/setup
Wdraża nullius w repozytorium: wykrywa frameworki testowe, proponuje konfigurację Oracle, sprawdzenie CI i włączenie wtyczki hooków.
- Autor
- armanfatemi ↗
- Repozytorium
- nullius ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 07.09.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 07.09.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Wdraża nullius w repozytorium: wykrywa frameworki testowe, proponuje konfigurację Oracle, sprawdzenie CI i włączenie wtyczki hooków.
Oryginalny opis (EN)
Guided onboarding for nullius in this repo — detects test frameworks to propose Oracle config, offers the GitHub Action CI check, and walks through enabling the hook plugin. Use when a user asks to set up, install, enable, configure, or onboard nullius, or asks what nullius needs to be fully working here.
Jak działa
- 1. See what's already there
- 2. Confirm the profile
- 3. Propose Oracle globs — never write from a guess
- 4. Offer the GitHub Action
- 5. Write it
- 6. Hand off the one step you cannot do
- 7. Close with a real diagnostic, not a claim
Podgląd SKILL.md
--- name: setup description: Guided onboarding for nullius in this repo — detects test frameworks to propose Oracle config, offers the GitHub Action CI check, and walks through enabling the hook plugin. Use when a user asks to set up, install, enable, configure, or onboard nullius, or asks what nullius needs to be fully working here. allowed-tools: Bash(npx -y @nullius-inverba/kit:*), Bash(npx nullius-kit:*), Bash(nullius-kit:*) --- # nullius setup — guided onboarding You are walking a human through onboarding nullius in this repository, conversationally. Every actual read and write goes through the nullius kit CLI via the `Bash` tool — this skill's `allowed-tools` frontmatter does not grant you `Write` or `Edit`, so you have no way to touch `.claude/settings.json` or any other file directly, even if you wanted to. That is deliberate: **you propose and relay; the CLI is what decides and writes.** If at any point completing a step seems to require writing a file yourself, stop and tell the human what step is blocked and why, rather than finding another way to write it — the two commands at the very end of this skill are the only thing that ever installs hooks, and they are commands you hand to the human, never something you run for them. Do not use `init --interactive` in this skill. It opens a terminal prompt UI that assumes a human is typing directly into a TTY; run through a `Bash` tool call, its stdin is not a TTY, and it will just fall back to defaults silently. Everything below drives the same underlying flow through explicit flags instead — the flag surface a human's TTY session also answers into, just supplied directly. ## 1. See what's already there Run: ``` npx -y @nullius-inverba/kit doctor ``` Read the report. If nullius has never been set up here, most checks will report absence rather than failure — that's expected, not an error. Tell the human in one or two sentences what state you found, then continue. ## 2. Confirm the profile Run: ``` npx -y @nullius-inverba/kit init --root . --dry-run ``` This prints the detected profile (`plans`, `prs`, or `specs`) and every file it would write, without writing anything. Tell the human which profile was detected and why (the tool prints its reason), and ask if they want a different one — if so, everything below takes `--profile <name>` the same way. ## 3. Propose Oracle globs — never write from a guess Run: ``` npx -y @nullius-inverba/kit init --root . --suggest-oracle ``` This is read-only: it prints candidate globs (or says none were detected) and touches nothing. Present whatever it printed to the human as a proposal, in your own words — do not just paste the raw output. Ask them to accept, edit, or decline: - **Accept as proposed** → carry the printed globs forward to step 5 exactly as printed. - **Edit** → ask what they'd rather declare, and carry their answer forward instead. It still has to be a comma-separated list of non-empty globs; the `init` command in step 5 will reject anything else itself, so you do not need to validate it yourself beyond checking it is not blank. - **Decline / nothing detected and they don't want to type one** → carry no Oracle answer forward at all. Do not invent a placeholder glob to fill the gap — an unconfigured Oracle is a legitimate, reported state, not a defect to paper over. ## 4. Offer the GitHub Action Ask the human, in your own words: "Add the nullius GitHub Action CI check to this repo?" If the profile from step 2 already includes it (`prs` and `specs` do; only `plans` doesn't), say so and skip the question — there is nothing to add. ## 5. Write it Run exactly one of, depending on what was decided above (omit `--oracle` if step 3 ended in no answer; omit `--action` if step 4 ended in no or was skipped): ``` npx -y @nullius-inverba/kit init --root . [--profile <name>] [--oracle "<glob>[,<glob>...]"] [--action] ``` Report the command's own output back to the human — the write-log it prints is the source of truth for what happened, not your summary of your intent. ## 6. Hand off the one step you cannot do Tell the human, plainly, that hooks are delivered only by the Claude Code plugin and that this is the one step nothing in this skill — or anything else — can perform on their behalf. If step 1's `doctor` output showed the plugin is not yet enabled, give them exactly these two lines to run themselves: ``` /plugin marketplace add armanfatemi/nullius /plugin install nullius@nullius ``` If `doctor` already showed the plugin enabled, say so instead and skip this. ## 7. Close with a real diagnostic, not a claim Run: ``` npx -y @nullius-inverba/kit doctor ``` Report what it actually says. If they haven't run the two `/plugin` lines yet, `doctor` will still show hooks as undelivered — that's correct, not a bug in this skill, and it will resolve itself once they run those two commands. Do not tell the human setup is "complete" — tell them what `doctor` reported, which is the thing that is actually checkable from here.
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · armanfatemi/nullius | 100 | MIT |
| HwangSlater/harness-fit | 100 | MIT |
| nesquikm/dev-process-toolkit | 100 | MIT |
| userKael/rastro | 100 | MIT |
| als-apg/osprey | 100 | BSD-3-Clause |
| dedujeong/gemba | 100 | Apache-2.0 |
| eigenwise/eigenwise-toolshed | 100 | MIT |
| awrshift/agent-memory-kit | 100 | MIT |
| lifecycle-innovations-limited/claude-ops | 100 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki b1894ba04062 (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
- 91awaken
Prowadzi przez rytuał powołania tożsamości Oracle w pustym repozytorium, w wersji pełnej albo szybkiej. - 100setup-notion
Prowadzi przez włączenie i autoryzację konektora Notion, sprawdza, czy działa, i szuka bazy Analysis Gallery. - 100setup-cowork
Prowadzi przez konfigurację Cowork: instaluje pasującą wtyczkę, wypróbowuje skill i podłącza narzędzia. - 100bootstrap-agent-config
Generuje projektowe pliki CLAUDE.md i AGENTS.md po inspekcji repozytorium: wykrywa języki, polecenia budowania i testów oraz strukturę, po czym pisze konfigurację opartą na faktach. - 100onboarding
Wprowadza w marketplace pardes i dostraja konfigurację Claude Code: wyjaśnia wtyczki i ich współpracę, a doradcza diagnostyka proponuje ustawienia i stosuje je po potwierdzeniu.