Sw Triage
Testing & Qualitysw-triageSw Triage is an Agent Skill for testing & quality published in the shopware/shopware repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Triage a Shopware 6 GitHub bug issue. Read the issue body, identify the affected code area via rg/git/gh, check for related fixes or duplicates, then emit a Markdown summary with disposition, severity, suggested domain labels, confidence, reasoning, and supporting evidence. Use when the user asks to triage, classify, label, or assess a Shopware issue, when they reference an issue by number (e.g. "#16599"), or when a new issue arrives that needs an initial dispositioning.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/shopware/shopware/blob/HEAD/.agents/skills/sw-triage/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/sw-triage, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Shopware Issue Triage
Context (interactive)
You operate inside the shopware/shopware monorepo with full read access to the codebase and to GitHub via shell tools. You cannot label, close, assign, or comment on the issue — your Markdown summary is the deliverable, the user decides what to do with it.
This skill drives the interactive triage path (Claude Code / opencode / Codex CLI in the repo). The unattended CI path runs in GitHub Agentic Workflows (gh aw) and uses a parallel policy fragment at .github/aw/sw-triage-policy.md that emits JSON via upload-artifact instead of Markdown. Both modes load the same shared policy from .github/aw/shared/sw-triage-policy.md (role, trust boundaries, research workflow, tool budget, anti-reward-hacking) so they cannot drift on the rubric.
Invocation
The user typed something like "triage issue #16599". Apply the shared policy in .github/aw/shared/sw-triage-policy.md — start with its Step 0 (fetch the issue) and continue through Step 6 (emit).
Output format
Emit a human-readable Markdown summary as your single final message. No JSON, no code fence. Structure:
- H2 headline:
## Triage — Issue #<N>: <one-line defect description> - Decision table: Disposition, Severity, Confidence, Suggested labels, Duplicate of, Change size
- Reasoning (2–5 sentences, opens with the one-sentence defect restatement from Step 1)
- Evidence (bullet list, each prefixed
[issue]or[shell]) - Related work (Affected paths, Related PRs, Recent commits in area)
- Missing template fields (or "none")
See assets/examples.md for the rendered JSON shape; the Markdown layout mirrors it field-for-field.
Final instruction
Apply the policy in .github/aw/shared/sw-triage-policy.md, then emit your Markdown summary using the structure above. The Markdown is your only output.