bugfix
/bugfix
Naprawia błąd w kodzie ABAP na podstawie zrzutu ST22 lub logu, śledząc łańcuch wywołań do przyczyny.
- Autor
- marian-kamenistak ↗
- Repozytorium
- sap-ai-skills ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 09.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 09.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Naprawia błąd w kodzie ABAP na podstawie zrzutu ST22 lub logu, śledząc łańcuch wywołań do przyczyny.
Wyzwalacze z opisu
Oryginalny opis (EN)
Fix a bug in SAP ABAP code from a Jira bug, an ST22 short dump, an error message or an application log, by tracing the call chain back to the cause, proving it with the code, writing a test that reproduces it and then the smallest fix. Use when someone shares a short dump (runtime error, exception, CX_ class), says "this report crashes", "fix this dump", "wrong value after save", "job fails every night", "users get error message X", "why does this BAdI fail", or a production incident lands in DEV for a fix. Not for unclear new requirements (use spec-grill), not for cleanup without a defect (us
Jak działa
- When to use
- When not to use
- How you reach the system
- Workflow (a person approves at each gate)
- Rules
- Output
Podgląd SKILL.md
---
name: bugfix
description: Fix a bug in SAP ABAP code from a Jira bug, an ST22 short dump, an error message or an application log, by tracing the call chain back to the cause, proving it with the code, writing a test that reproduces it and then the smallest fix. Use when someone shares a short dump (runtime error, exception, CX_ class), says "this report crashes", "fix this dump", "wrong value after save", "job fails every night", "users get error message X", "why does this BAdI fail", or a production incident lands in DEV for a fix. Not for unclear new requirements (use spec-grill), not for cleanup without a defect (use refactor).
---
# bugfix
Find the cause first, prove it, then fix it with the smallest change and a test that stops it
coming back. Guessing at a fix from the dump text alone is how bugs get moved instead of fixed.
A person confirms the cause before any code changes.
## When to use
- A short dump in ST22, with a runtime error such as `ITAB_DUPLICATE_KEY`,
`GETWA_NOT_ASSIGNED`, `CONVT_NO_NUMBER`, or an uncaught exception class
- A wrong result reported in Jira ("the total is off", "status not updated")
- A batch job that fails, with its job log or application log (SLG1)
## When not to use
- The "bug" is a missing feature or a changed requirement: use spec-grill
- You only need to understand the code: use reverse-engineer
- SAP standard code is at fault: search SAP notes and raise an incident; do not modify standard
## How you reach the system
Say which setup you are using before step 1.
- **ABAP MCP server that can write (DEV only).** Read the dump (if the server offers it), code,
DDIC and where-used live. Write the failing test and the fix in DEV in an allowed package, run
ATC and ABAP Unit.
- **Read-only ABAP MCP server.** Read and run tests over MCP. Hand the test and the fix to a
person who applies them in ADT; then run the checks.
- **Files exported with abapGit.** Work on the export plus the dump text a person pastes. Write
the test and fix as files; a person imports them and runs the checks.
The dump text must reach you without personal data. Variable values in ST22 can hold names,
addresses or bank data: ask the person to remove or mask them before pasting.
## Workflow (a person approves at each gate)
**1. Read the evidence.** From the dump: runtime error, exception class, program, include, line,
the "Active Calls/Events" call stack, and the variable values relevant to the failing statement.
From a wrong result: the steps, the input, expected vs actual. Restate the bug in one sentence.
**2. Trace back.** Start at the failing line and walk the call chain back toward the entry point.
At each hop, ask where the bad value or state came from. Read the where-used if a shared object
is involved.
**3. Name the cause, then STOP.** Describe the cause and show the lines of code that prove it.
If there are several candidates, rank them with evidence. Give the debugger breakpoints that
would confirm it. A person confirms the cause before you continue.
**4. Reproduce with a test.** Write an ABAP Unit test that fails on the current code for exactly
this bug, using test doubles for the database (`CL_OSQL_TEST_ENVIRONMENT`) and dependencies. If
the code cannot be tested, make the smallest testability change first and say so.
**5. The smallest fix.** Change only what the cause needs. Activate, run the new test, the
existing tests and ATC. All must pass. Check the where-used: does the same pattern exist in other
places? Report them; do not fix them silently in this transport.
**6. Show the diff, then STOP for review.** A person reviews the fix and the test.
**7. Transport and record.** Assign the fix and the test to one transport. Write the cause, the
fix and the test into the Jira ticket. A person releases the transport.
## Rules
- No fix before a confirmed cause.
- Never debug or test in production. Reproduce in DEV with invented test data.
- Never read table contents to "see the bad record". Ask a person to describe the data pattern
(for example "two items with the same key") without real values.
- The smallest change that fixes the cause. Refactoring goes to a separate transport.
- A person releases the transport.
## Output
1. The bug in one sentence
2. Call chain from entry point to the failing line
3. The cause with code evidence, confirmed by <person>
4. The failing test, then passing
5. The fix as a diff, with ATC and test results
6. Same pattern elsewhere (if any)
7. Jira note: cause, fix, test, transport
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · marian-kamenistak/sap-ai-skills | 100 | MIT |
| qwenlm/qwen-code | 100 | Apache-2.0 |
| grimoire-rs/grimoire | 100 | Apache-2.0 |
| duthaho/skillhub | 100 | MIT |
| alibaba/mnn | 95 | Apache-2.0 |
| 4typen/skills | 90 | MIT |
| kangig94/coral | 90 | MIT |
| woliveiras/baseline | 86 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki e41f54e0d3c6 (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
- 90reverse-engineer
Wyjaśnia działanie istniejącego kodu ABAP i przyczyny błędów, śledząc punkt wejścia, łańcuch wywołań i przepływ danych. - 90implement
Buduje nową funkcję SAP ABAP ze specyfikacji do aktywowanego kodu, z testami ABAP Unit dla każdego kryterium. - 100dump-triage
Segreguje krótkie zrzuty ABAP (ST22): wyświetla ostatnie, grupuje według przyczyny źródłowej, czyta rozdziały zrzutu, pobiera kod w miejscu przerwania i tworzy raport z poprawką dla każdej rodziny zrzutów. - 90pm-bugfix
Prowadzi naprawę błędu od odtworzenia i analizy przyczyny, przez minimalną poprawkę, po testy regresji. - 100bug-investigation
Prowadzi błąd od zgłoszenia do testu regresyjnego: fakty, bezpieczne odtworzenie i poprawka przyczyny. - 100implementing-bug-fix
Naprawia błąd ze zgłoszenia GitHub lub wybiera kolejny potwierdzony błąd, odtwarza go, testuje i tworzy gotowy PR.