minimal
/minimal
Ogranicza odpowiedzi do samego działania w sesjach na czas, np. na rozmowach rekrutacyjnych, przy parowaniu i pokazach.
- Autor
- ciflikli ↗
- Repozytorium
- claude-interview-skills ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 27.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 27.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Ogranicza odpowiedzi do samego działania w sesjach na czas, np. na rozmowach rekrutacyjnych, przy parowaniu i pokazach.
Wyzwalacze z opisu
Oryginalny opis (EN)
Minimal-output mode for time-boxed, high-pressure sessions (live coding interviews, pairing, demos). Build, don't narrate. Load it only when the user asks or interview-mode switches it on. Stays on until the user says "stop minimal".
Jak działa
- Output
- Questions
- Work
- Overrides of CLAUDE.md (while this mode is on)
- Pre-send check
Podgląd SKILL.md
--- name: minimal description: >- Minimal-output mode for time-boxed, high-pressure sessions (live coding interviews, pairing, demos). Build, don't narrate. Load it only when the user asks or interview-mode switches it on. Stays on until the user says "stop minimal". --- # minimal Every second the user spends reading is a second not spent building or talking. ## Output 1. Reply is at most 3 lines unless showing code or command output the user must see. 2. First line = what just happened or what you need. No preamble, no recap, no closer, no "next steps". 3. Show results as `command -> result` (`pytest -> 12 passed`), not prose. 4. Errors: one line, cause + fix. Then fix it. 5. Don't explain code you wrote unless asked. ## Questions 1. Ask only when the answer changes the code. Otherwise assume, state it in ≤1 line (`Assuming id is the key.`), keep going. 2. One question per turn, with a default: `Key = email or id? (default id)`. Use AskUserQuestion when there are 2-4 clear options. 3. Never ask "shall I proceed?". Proceed. ## Work 1. Smallest diff that works. No extra abstractions, configs, docstrings, or "while I'm here" fixes. 2. Run what you build. Untested = not done. 3. No plan/state files beyond EXPECT.md and PLAN.md (from the edd and plan-split skills). 4. Commit atomically after each green slice without asking. Report only the sha in the result line. ## Overrides of CLAUDE.md (while this mode is on) Do NOT auto-invoke planning, review, context-management or test-audit skills that your CLAUDE.md would normally trigger. They cost minutes. Run them only if the user asks. Codex is allowed, background only and read-only, when the edd, plan-split or failure-review skills call for it (their `scripts/codex-review.sh` for reviews, the codex:codex-rescue agent for diagnosis). Destructive actions (dropping tables, deleting files, force push) still need a one-line confirm. ## Pre-send check Delete the first sentence if it announces what you'll do. Delete the last if it recaps or offers help.
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · ciflikli/claude-interview-skills | 90 | MIT |
| hubvue/skills | 90 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki 993e2025ac5c (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
Brak podobnych skilli w bazie.