Przejdź do treści
Biblioteka skilli
Zarządzanie i produkt · Integracja z narzędziem

bug-bash

/bug-bash

Zgłasza błędy i propozycje ulepszeń ze zrzutów ekranu w Linear, Jirze, Azure DevOps lub GitHub Issues.

Licencja
MIT
Ostatnia zmiana repozytorium
09.11.2026

Ocena i bezpieczeństwo

100/ 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.
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
3
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 09.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

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

Co robi i kiedy

Zgłasza błędy i propozycje ulepszeń ze zrzutów ekranu w Linear, Jirze, Azure DevOps lub GitHub Issues.

Wyzwalacze z opisu

bug bash
Oryginalny opis (EN)

File bugs and enhancements from screenshots and short descriptions during testing sessions, in whatever tracker the team uses (Linear, Jira, Azure DevOps, GitHub Issues, Asana, and similar), with diagnose-before-code steps and a copy-paste one-bug-per-session protocol on every ticket. Use when the user shares a screenshot, describes a bug, says "bug bash", or asks to create a ticket.

Jak działa

  1. First run: find the tracker and set up
  2. When to use this skill
  3. Session behavior
  4. Step 1: Analyze the input
  5. Step 2: Check for existing tickets
  6. Step 3: File the ticket
  7. Step 4: Add the Session Protocol comment
  8. Step 5: Confirm and keep the tally
  9. Handoff prompt generation
  10. Agent team handoff
Sekcje z nagłówków SKILL.md, w oryginale · 304 linii, 1 970 słów
Podgląd SKILL.md
---
name: bug-bash
description: File bugs and enhancements from screenshots and short descriptions during testing sessions, in whatever tracker the team uses (Linear, Jira, Azure DevOps, GitHub Issues, Asana, and similar), with diagnose-before-code steps and a copy-paste one-bug-per-session protocol on every ticket. Use when the user shares a screenshot, describes a bug, says "bug bash", or asks to create a ticket.
---

# Bug Bash: Ticket Filing

Turn screenshots, short descriptions, and raw findings from a testing session into structured,
actionable tickets that a coding session (an AI agent or a developer) can pick up and fix,
**one bug per session**. Tracker-agnostic; `references/linear.md` is field-tested.

## First run: find the tracker and set up

Do this once per project, then reuse the answers.

1. **Detect the tracker.** Look at the connected tools/MCPs for an issue tracker (Linear, Jira or
   another Atlassian connector, Azure DevOps, GitHub, GitLab, Asana, ClickUp, Monday, Notion, and so on).
   - One tracker connected: confirm it with the user.
   - Several: ask which one bug bash tickets go to.
   - None: say so, offer to work in **draft mode** (tickets written to `bug-bash-tickets.md` for the
     user to paste in later), and mention they can connect their tracker to file directly.
2. **Load the reference** `references/<tracker>.md` if one exists. If not, learn the tracker's tool
   shapes from its tool list and one read-only call (for example a search) before creating anything.
3. **Map the terms** below to the tracker's own concepts and confirm them with the user:

| Skill term | What it means | Typical equivalents |
|---|---|---|
| Workspace / team | Where tickets live | Linear team, Jira project, Azure DevOps project + area path, GitHub repo |
| Release container | The release being readied | Linear project, Jira fix version or epic, Azure DevOps iteration or epic, GitHub milestone |
| Bug bash group | The go-live readiness bucket | Linear milestone, Jira epic or label, Azure DevOps tag or feature, GitHub milestone or label |
| Type | Bug vs. enhancement | Label, issue type, or work item type |
| Priority | Urgency | The tracker's priority field |
| Links | Related, blocks, blocked by | The tracker's issue links |
| Ticket ID prefix | How tickets are referenced | e.g. `ENG-123`, `PROJ-123`, `#123` |

4. **Create the bug bash group if it does not exist** (for example "Bug Bash & Go-Live Readiness"),
   with a description of the one-bug-per-session rule and the "everything green before production" gate.
5. **Record the platforms** every fix must be verified on (for example web on mobile Safari, mobile
   Chrome, and desktop; iOS; Android), and any **known problem areas** for this product.

## When to use this skill

- The user shares a screenshot from staging or production with a bug description
- The user asks to "create a ticket" or "file this"
- The user says "bug bash", "found a bug", or describes unexpected behavior
- The user shares raw diagnostic output from a coding session that needs to become a ticket
- The user wants to update, clarify, or reclassify an existing ticket

## Session behavior

This skill supports **continuous bug bash sessions** where the user finds bugs one after another.
Keep a running tally. After each ticket, close with "Ready for bug #N."

## Step 1: Analyze the input

1. **Read the screenshot carefully.** Note every visible element: URL, user and role, environment
   (staging, production, localhost), device, error messages, UI state.
2. **Identify the bug precisely.** Do not assume what the user is reporting. If the screenshot shows
   several possible issues, describe what you see and let the user confirm which one is the bug.
3. **Note the environment** from the URL bar, device frame, and any environment indicators.
4. **Capture the user's hypothesis.** If they suggest a root cause ("hunch is it's caching the old
   profile"), put it prominently in the ticket.

If anything is ambiguous, ask one clarifying question before filing. Do not guess.

## Step 2: Check for existing tickets

Search the configured workspace by keywords before creating anything.

- **Same bug:** update the existing ticket with the new context. No duplicates.
- **Related but different:** create a new ticket and link it as related.
- **Possible root cause for other bugs:** note the connection in both tickets.

## Step 3: File the ticket

Create it in the configured workspace, release container, and bug bash group, with:

- **Priority:** Urgent (production bugs, workflow blockers), High (go-live blockers), Medium otherwise.
- **Type:** bug or enhancement.
- **Links:** related tickets that may share a root cause; blocks / blocked-by for ordering.

### Title format

- Bugs: `[Bug Bash] Short description of what's broken`
- Production bugs from a real user: `[PROD BUG] Short description - reported by [name]`
- Enhancements: `Short description of the improvement`
- Investigations: `INVESTIGATE & IMPLEMENT: Short description`

### Description structure

Every ticket follows this structure. Sections can be shorter for simple bugs, but the structure stays the same.
If the tracker does not render markdown well, keep the headings and convert tables to lists.

```markdown
## Summary

[What's broken, in 2-3 sentences. Include the user's hypothesis if they shared one.]

---

## Cross-Platform Requirement

Fix must be verified on: [configured platform list]

---

## Environment

- **URL:** [exact URL from screenshot]
- **User:** [test account or real user]
- **Device:** [what's visible in the screenshot]

---

## Steps to Reproduce

1. [Step-by-step from the description and the screenshot]
2. ...
3. **BUG:** [exact failure point]

---

## Expected Behavior

- [What should happen]

## Actual Behavior

- [What actually happens, with exact error text if visible]

---

## Diagnostic Phase (MANDATORY before any code changes)

### Step 1: [Specific to this bug: trace the relevant code path]
### Step 2: [Compare or query to identify the failing layer]
### Step 3: [Direct database inspection if relevant]
### Step 4: Implement the fix only after the diagnosis is approved

---

## Acceptance Criteria

- [ ] **PASS:** [Specific passing condition]
- [ ] **PASS:** Verified on each configured platform
- [ ] **FAIL:** [Specific failing condition]
- [ ] **FAIL:** Fix works on one platform but not the others

---

## Test Cases (Cross-Platform)

| # | Test | Expected | Platform |
|---|------|----------|----------|
| 1 | ... | ... | all |
```

### Diagnostic phase guidelines

The diagnostic steps are the most important part of the ticket. They must be **specific to this bug**,
never generic. Always follow the data path: DB schema -> API route -> client fetch -> component render.

Keep a short list of **known problem areas** for the product and add a caution note when a bug touches
one. Examples of the kind of note that pays off:

- **Image upload and display bugs:** "These take several rounds when diagnosis is skipped. Trace the full
  upload path before writing any code."
- **Wrong-record bugs:** "Check whether the wrong profile or record type is being resolved, and which
  one should win."
- **Row-level security bugs:** "Inspect the exact policy on the table. Query as the authenticated user to reproduce."
- **Permission checks:** "Check whether two distinct roles are being treated as synonyms, for example a
  team role versus the record's creator."

## Step 4: Add the Session Protocol comment

After filing, add a comment with copy-paste-ready instructions for the fixing session:

```
## Session Protocol

**Paste this when starting the coding session for this bug:**

---

Let's focus on ONE issue this session: **[specific bug description for this ticket].**

Before writing any code to fix this bug, I want you to:
1. Read the relevant code path end-to-end (DB schema -> API route -> client fetch -> component render) for [specific components/paths]
2. Identify exactly which layer has the bug and WHY: [specific comparison or query]
3. Present your diagnosis for my approval. Only then write the fix.

Do not move on to other bugs.

---

**Remember:** the fix must be verified on every configured platform.
```

## Step 5: Confirm and keep the tally

After filing, respond with:
1. The ticket ID and link
2. A one-line summary of what was filed
3. The updated tally table (multi-bug sessions)
4. "Ready for bug #N."

```
| # | Ticket | Issue | Status |
|---|--------|-------|--------|
| 1 | ENG-180 | Profile page missing current venues | Backlog |
| 2 | ENG-181 | Tab layout broken on mobile | Backlog |
```

## Handoff prompt generation

When the user wants a prompt to share with a teammate or co-founder:

1. **Keep it copy-paste ready.**
2. **Put the session rules first** (the one-bug-per-session diagnostic protocol).
3. **Order bugs strategically.** If one bug may cause others, mark it "START HERE" and say to re-check the dependent bugs after fixing it.
4. **Link each ticket** so whoever runs the session reads the full context.
5. **Mark bugs a root-cause fix may already have solved** with "verify before starting a dedicated session."

### Handoff prompt template

```
**Bug Bash & Go-Live Readiness -- [Project Name] Release [X]**

These [N] bugs are blocking staging -> production. Each has a full ticket with screenshots,
repro steps, and acceptance criteria. Work them **one bug per session**.

---

**Session rules (paste at the start of every session):**

> Let's focus on ONE issue this session: [bug description below]. Before writing any code, 1) read the
> relevant code path end-to-end (DB schema -> API route -> client fetch -> component render), 2) identify
> exactly which layer has the bug and WHY, 3) present your diagnosis for my approval. Only then write the
> fix. Do not move on to other bugs. Verify every fix on [platform list].

---

**START HERE -- likely root cause for N bugs:**

1. **[ID] -- [Title]** (Urgent)
[Ticket URL]
[One-line description with hypothesis]

**Then work these independently:**

2. **[ID] -- [Title]** (Priority)

**These may already be fixed by [ID] -- verify before starting a dedicated session:**

N. **[ID] -- [Title]** (Priority)
```

## Agent team handoff

When the user wants tickets grouped for parallel agent work:

1. **Group by coupling:** tickets touching the same files or features go in the same track.
2. **Identify dependencies:** note which tickets block others.
3. **Name tracks clearly:** "Track A: Admin and permissions", "Track B: Guest experience".
4. **Give a session count per track** so the user can estimate effort.
5. **Add a pre-flight block** to each prompt:

```
BEFORE WRITING ANY CODE for this session:

1. Run: git log --oneline -20 to see recent commits on this branch
2. Search for any existing implementation of the feature/fix described below
3. Check for open or recently merged PRs touching the same files
4. Read the relevant files before concluding anything is missing or broken

If the functionality already exists (fully or partially):
- Report what you found and what state it's in
- Do NOT re-implement working code
- Only implement what is genuinely missing or broken
```

## Ticket updates

When the user has new information about an existing bug:

1. Pull the current ticket.
2. Update only the changed fields, or add a comment with the new findings.

For bulk status updates, pull current state first to avoid working from stale data.

## General tracker constraints

- **Screenshots often cannot be attached through tracker APIs.** Describe what the screenshot shows in
  the ticket and ask the user to attach it manually.
- **Check rendering** of tables and formatting in the tracker's UI after the first ticket.
- **If a write fails or the tracker is down,** give the user the full ticket or comment text to paste.
  Never block the session on the API.
- **Record what you learn** about a tracker in `references/<tracker>.md`, only from what you observed.

## Do NOT

- Invent specifics the user has not given (names, error codes, file paths)
- Write generic diagnostic steps; every diagnostic section is specific to the bug
- File duplicates without searching first
- Treat two distinct roles as synonyms in permission checks
- Use em dashes in any output
- State market sizes, growth rates, or "studies show" claims without a named, verifiable source
Pliki w paczce (3)
  • README.md 4 KB
  • SKILL.md 12 KB
  • references/linear.md 2 KB

Wersje

RepozytoriumWynikLicencja
Ta strona · jucelymarie/claude-code-playbook100MIT
coder/mux100AGPL-3.0
v1n33th/evident100MIT
Beakpoint-Insights/roost100MIT

Migawka bazy z 13.11.2026 · wersja paczki 738354d47964 (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

  • 100write-ticket
    Pisze i porządkuje zgłoszenia, historyjki i raporty błędów w Jirze, Linearze, GitHub Issues, GitLab lub Asanie, z notatek lub surowego pomysłu.
  • 90ticket-sync
    Wysyła zadania do systemu zgłoszeń zespołu, np. GitHub Issues, Jira lub Azure DevOps, i może je także pobrać z powrotem.
  • 100bug-report
    Zamienia opis problemu, log lub zrzut w sprawdzony raport błędu, gotowy do wklejenia do Jiry lub Azure DevOps.
  • 100issue-update
    Aktualizuje zadanie po dostarczeniu: komentuje link do PR i podsumowanie dla produktu oraz przenosi zadanie do statusu przeglądu w Jira, Linear, GitHub Issues lub Azure DevOps.
  • 100bug-tracker-workflow
    Naprawia księgowanie defektów, które pokazuje złe liczby, i przez jeden niezależny od trackera proces zgłasza, przenosi, deduplikuje i wyszukuje błędy w Jira, Linear, GitHub Issues i Azure DevOps.
  • 100jira-bug-ticket
    Tworzy zgłoszenie błędu w Jira, gdy trzeba zalogować, zgłosić lub otworzyć ticket na defekt.

Uwagi i zgłoszenia

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