Programowanie · Testowanie
edge-case
/edge-case
Sprawdza ostatnią pracę pod kątem przypadków brzegowych, które wypisał użytkownik.
Dla kogo
- Autor
- yuri-xyz ↗
- Repozytorium
- skills ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 03.11.2026
Ocena i bezpieczeństwo
96/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność21 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 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
- 2
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Przestarzałość
Brak sygnałów. Repozytorium zmienione 03.11.2026, bez odwołań do wycofanych modeli i API.
Licencja
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Sprawdza ostatnią pracę pod kątem przypadków brzegowych, które wypisał użytkownik.
Oryginalny opis (EN)
Use when the user lists edge cases and wants recent work checked against them.
Jak działa
- Objective
- Protocol
- Trivial Versus Decision
- Boundaries
- Output Shape
Sekcje z nagłówków SKILL.md, w oryginale · 93 linii, 700 słów
Podgląd SKILL.md
---
name: edge-case
description: Use when the user lists edge cases and wants recent work checked against them.
disable-model-invocation: true
---
# Edge Case
## Objective
The user is pointing at one or more possible edge cases. Check whether the
agent's most recent work actually handles each one, then close the gaps:
fix the ones with an obvious answer, settle the ones that need a technical
decision with `/decide`, and confirm only the ones that need a human decision
with the user before changing them.
Treat the user's cases as hypotheses to verify, not as confirmed bugs. A case
may already be handled, may be impossible given the surrounding invariants,
or may expose a real gap.
## Protocol
1. Restate each edge case as a concrete scenario: the input, state, or
sequence, and what correct behavior should be. Split compound cases.
2. Identify the recent work in scope: the changed files, the plan or task
being implemented, and the code paths those changes own.
3. For each case, trace it through the real code. Read the implementation,
callers, and invariants; run an existing test or a narrow check when that
is cheaper and more decisive than reading.
4. Classify each case as exactly one of:
- **Covered**: the code already handles it. Cite the file and line or the
test that proves it.
- **Not reachable**: an upstream invariant, type, or validation prevents it.
Cite what prevents it.
- **Trivial gap**: unhandled, and the correct behavior is obvious from the
code, the surrounding conventions, or the user's stated intent.
- **Decision gap**: unhandled, and closing it requires a choice about
behavior, architecture, data format, compatibility, scope, or risk.
5. Fix every trivial gap directly at the owner of the behavior, with a
focused test or assertion when the project has a test surface for it.
Keep fixes flat per `/root-cause`; do not patch symptoms downstream.
6. Classify each decision gap as technical or human by `/decide`'s "Human
versus technical decisions". Settle each technical gap with `/decide` as a
delegated decision, then fix it like a trivial gap and report the choice
and its reason so the user can override it.
7. For human gaps, use `/interview`: one question per gap, with `/decide`'s
recommendation first and marked `(Recommended)`. Do not implement a human
gap before the user answers; after answers arrive, implement the confirmed
approaches and validate.
## Trivial Versus Decision
Fix without asking when the answer has one clearly correct shape, such as:
- a missing bounds, empty, null, or zero check with an obvious result;
- an unhandled error that should propagate like its neighbors already do;
- escaping, quoting, or encoding that the project already models elsewhere;
- a missing test for behavior that is already correct.
It is a decision gap when a reasonable engineer could pick differently, such
as:
- choosing between rejecting, clamping, defaulting, or silently accepting input;
- changing a public interface, data format, persisted state, or migration;
- adding a new component, dependency, state, retry, or fallback path;
- trading correctness against compatibility, performance, or simplicity;
- widening the work's scope beyond what the user asked for.
When unsure which side a case falls on, treat it as a decision gap. Most
decision gaps are technical and go to `/decide`; ones that change a public
interface, persisted data, or the work's scope, or that rest on product
intent, are usually human.
## Boundaries
- Stay on the cases the user raised plus directly adjacent cases the trace
reveals. Mention other discoveries instead of fixing them.
- Do not invent an interview when no human decision gap remains.
- Do not weaken checks, suppress errors, or add fallbacks to make a case
"pass"; follow `/fallbacks` when a case tempts a default value.
## Output Shape
Report once per invocation, before any interview question:
```markdown
| Edge case | Status | Evidence or action |
| --- | --- | --- |
| <scenario> | Covered | `path/to/file:42` rejects it before use |
| <scenario> | Fixed | <what changed>, tested by `<test>` |
| <scenario> | Decided | <choice> (`/decide`: <one-line reason>), tested by `<test>` |
| <scenario> | Needs decision | <the human choice to make> |
Validation: <checks run and results>
```
Then start `/interview` for the human decision gaps, if any remain.
Pliki w paczce (2)
- SKILL.md 4 KB
- agents/openai.yaml 1 KB
Wersje
W bazie nie ma innych skilli o nazwie edge-case.
Migawka bazy z 13.11.2026 · wersja paczki 7fe1c3b7ae97 (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
- 100ctx-code-review
Przegląda zmiany w kodzie, diff lub PR pod kątem poprawności, przypadków brzegowych i zgodności z konwencjami. - 100code-review
Przegląda kod i diffy pod kątem błędów, regresji, przypadków brzegowych, wyścigów i wycieków przed scaleniem. - 100bugmagnet
Wykrywa błędy i luki w pokryciu testami modułu kodu, pisze testy przypadków brzegowych i prowadzi testy eksploracyjne wskazanego pliku. - 100resilience-review
Przegląda kod pod kątem przypadków brzegowych, obsługi błędów, stanów i dopracowania, zanim zmiana pójdzie dalej. - 100qa-review
Ocenia gotowość funkcji do wydania: weryfikacja kryteriów akceptacji, lista przypadków brzegowych, ryzyko regresji i uzasadniona decyzja o wydaniu. - 90Code Review
Pomaga w przeglądzie diffu lub pull requestu pod kątem przypadków brzegowych, bezpieczeństwa, testów i czytelności.