Przejdź do treści
Biblioteka skilli
Programowanie · Przegląd i audyt

preflight

/preflight

Ocenia uczciwie stan pracy przed wysyłką: wskazuje, co jest dobre, niedokończone, błędne lub do poprawy, a z opcją naprawy poprawia to, co się da.

Repozytorium
claude-plugins ↗
Licencja
MIT
Ostatnia zmiana repozytorium
20.10.2026

Ocena i bezpieczeństwo

90/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść25 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
1
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 20.10.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co robi i kiedy

Ocenia uczciwie stan pracy przed wysyłką: wskazuje, co jest dobre, niedokończone, błędne lub do poprawy, a z opcją naprawy poprawia to, co się da.

Wyzwalacze z opisu

how happy are youhow do you feel about thispreflight checkhonest assessmentare you satisfiedwhat. Surface what
Oryginalny opis (EN)

Honest self-assessment of the current state of work before shipping. Use when the user invokes the preflight skill or asks "how happy are you", "how do you feel about this", "preflight check", "honest assessment", "are you satisfied", or "what's the state of things". Surface what's genuinely good, incomplete, buggy, or improvable. Pass --fix to fix actionable issues and repeat the assessment until no autofixable issues remain.

Jak działa

  1. Modes
  2. How to reflect
  3. NOW vs LATER
  4. Output
  5. After
Sekcje z nagłówków SKILL.md, w oryginale · 71 linii, 814 słów
Podgląd SKILL.md
---
name: preflight
description: Honest self-assessment of the current state of work before shipping. Use when the user invokes the preflight skill or asks "how happy are you", "how do you feel about this", "preflight check", "honest assessment", "are you satisfied", or "what's the state of things". Surface what's genuinely good, incomplete, buggy, or improvable. Pass --fix to fix actionable issues and repeat the assessment until no autofixable issues remain.
---

# Preflight: honest assessment before shipping

Stop and honestly assess: how happy are you with the current state of the work?

Not "are you happy" (binary, invites a polite yes). "HOW happy" (graduated, demands nuance). This is your chance to surface everything you've been holding back, rationalizing, or planning to mention later.

## Modes

Usage: `preflight [--fix]` (use the host's skill invocation syntax).

- **Without `--fix`**: assess and report. Wait for the user to choose what to act on.
- **With `--fix`**: assess, fix, verify, and repeat. The flag authorizes fixes within the current task, not new scope, destructive actions, or bypassing permissions.

## How to reflect

Review what actually exists. Read files, check the diff, examine the produced artifacts. Don't assess from memory.

Self-assessment has a built-in blind spot: you're grading your own work. For high-stakes ships, follow `second-opinion` if available and fold any findings into the assessment.

Then assess across these dimensions:

### Completeness
Is everything implemented or written? Were any requirements quietly dropped, or items deferred without discussion?

### Correctness
Does it actually do what it claims? Any known bugs, edge cases, or unverified assumptions?

### Quality
Is this clean, correct, and elegant, or were corners cut? Would you be proud to show this to a sharp peer who'd notice problems?

### Verification
Has it been properly checked (tests, manual usage, rendered output, peer-read, whatever's appropriate)? Anything that should have been verified but wasn't?

### Loose ends
What's unfinished or deferred? Any risks or fragilities introduced?

## NOW vs LATER

Default concrete issues in the current work to **NOW**. Bugs, missing requirements, verification gaps, stale docs, and quality problems a sharp reviewer would flag belong here. Being preexisting, small, tedious, or more work than expected is not a reason to defer them.

**LATER is the exception**: an independent enhancement not needed to ship the current work correctly, or something the user explicitly chose to defer. Explain why leaving each item does not compromise this ship; for user-directed deferrals, record the decision and any remaining risk. Don't turn unfinished work into a follow-up just to finish the session.

A NOW item needing access, approval, or a user decision is **blocked NOW**, not LATER. Separate urgency from whether you can fix it yourself.

## Output

Be direct. No diplomatic softening.

**Overall**: a genuine feeling, not a score. *"I'm happy with the core logic but uncomfortable with the error handling"* beats *"7/10."*

**What's good**: briefly. Don't pad to soften the bad news.

**What needs attention NOW**: specific remaining items to fix before shipping. Include enough detail to act on: file paths, line numbers, what's wrong, what to do about it. For blocked items, state the blocker and the exact input, access, or approval needed.

**What should be filed for LATER**: only items that meet the deferral rule above. For each: title + why it matters + why it can wait. An empty list is fine.

## After

**Without `--fix`**: present the assessment and wait. The user decides what to act on. Don't preemptively fix things or open tickets.

**With `--fix`**, run this loop in the current conversation:

1. **Assess** the actual artifacts across all five dimensions above. Validate findings before acting; if evidence refutes one, say why rather than fixing a non-issue.
2. **Fix** every autofixable NOW item: a confirmed issue with a clear correction you can make within the current task and permissions. Fix the cause, not just the symptom. Keep blocked items visible and continue with independent fixes; don't pause for approval of already-authorized work.
3. **Verify** the fixes with the relevant tests, checks, or direct use of the artifact. A change isn't fixed until verification supports it. Failed or unavailable verification remains NOW; repair what you can and name any real blocker.
4. **Run preflight again with `--fix`** on the updated work, starting at the full assessment, not just the previous findings list. Re-read the artifacts and diff with tools; remembering your edits is not a fresh assessment. Do this even when known remaining issues are blocked. Include newly exposed issues and regressions. Repeat after every fix batch; a passing test run alone is not the exit condition.
5. **Stop only when a fresh assessment finds no autofixable NOW issues.** Don't stop merely because one fix pass is done, impose an arbitrary pass count, or relabel unresolved items as LATER to exit. If a repair stalls or the same failure recurs, investigate the cause and change approach; if you cannot proceed, record the attempted fixes, evidence, and concrete blocker rather than retrying blindly or claiming success.

Keep a concise record of findings, fixes, and verification across passes so nothing disappears. At the end, present the assessment of the **final state**, plus what was fixed and how it was verified. If blocked NOW items remain, say the work is still blocked, not clean or ready to ship. Don't open tickets unless asked.

Wersje

RepozytoriumWynikLicencja
Ta strona · pandysp/claude-plugins90MIT
andreibesleaga/gabbe100CC-BY-4.0
pathmodeio/claude-plugin100MIT
aradmanamnaoon/agent-guardrails100MIT
mo-cmyk/mazarbul100Apache-2.0
thilak15/preflight100MIT
manpreet171/slopguard100MIT
rune-kit/rune100MIT
nazmirket/skills90MIT

Migawka bazy z 13.11.2026 · wersja paczki 92fafb638da8 (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

  • 100workflow-preflight
    Sprawdza przed implementacją stan zdalny, istniejące pull requesty, konflikty gałęzi, autoryzację gh i bramki merge przy starcie zadania lub rozsyłaniu agentów.
  • 95git-preflight
    Uruchamia skonfigurowane w projekcie kontrole typów, lintu, formatu, bezpieczeństwa i testów przed commitem i proponuje naprawę tego, co nie przeszło.
  • 100verify-preflight
    Uruchamia izolowaną instancję Preflight z tego checkoutu i sprawdza panel, API HTTP, przechwytywanie hooków i koszty tokenów.
  • 90web-app-preflight
    Sprawdza gotowość aplikacji webowej do wdrożenia: widoczność, analitykę, podglądy linków, mobilny wygląd i strony 404, z wynikami PASS lub FAIL.
  • 100implement-plan-preflight
    Uruchamia fazę Preflight w workflow ITP i tworzy artefakty ADR oraz specyfikację projektową, gdy ktoś prosi o napisanie ADR.
  • 91project-preflight
    Sprawdza repozytorium, worktree, powłokę, terminal, GitHub CLI i uprawnienia przed startem pracy w projekcie.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.