gh-create-issue
Use when user wants to create a GitHub issue for the current repository. Must read and follow the repository's issue template format.
git clone --depth 1 https://github.com/CherryHQ/cherry-studio /tmp/gh-create-issue && cp -r /tmp/gh-create-issue/.agents/skills/gh-create-issue ~/.claude/skills/gh-create-issueSKILL.md
# GitHub Create Issue
Use this skill when the user requests to create an issue. Must follow the repository's issue template format.
## Workflow
### Step 1: Determine Template Type
Analyze the user's request to determine the issue type:
- If the user describes a problem, error, crash, or something not working -> Bug Report
- If the user requests a new feature, enhancement, or additional support -> Feature Request
- If the user is asking a question or needs help with something -> Questions & Discussion
- If the user describes engineering work with no user-facing behavior change (a follow-up deferred from a pull request, technical debt, refactoring, tooling or CI work) -> Engineering Task
- Otherwise -> Others
Engineering Task is not a GitHub issue form and is not offered to regular users. Only select it when the request clearly describes internal engineering work.
**If unclear**, ask the user which template to use. Do not default to "Others" on your own.
#### Eligibility Check (Engineering Task only)
Engineering Task is meant for maintainers and collaborators with write access. This is a request,
not an access control — nothing prevents anyone from calling `gh issue create` directly, and this
skill does not try to. The check below is here so that a contributor without write access learns
that early and gets pointed somewhere more useful.
Before collecting any information, read the authenticated user's permission level:
```bash
repo="$(gh repo view --json nameWithOwner --jq .nameWithOwner)"
user="$(gh api user --jq .login)"
permission="$(gh api "repos/$repo/collaborators/$user/permission" --jq .permission 2>/dev/null)"
case "$permission" in
admin | write) echo "eligible" ;;
*) echo "not eligible" ;;
esac
```
Only `admin` and `write` are eligible (`maintain` reports as `write`). Always compare the returned
value: on a public repository this endpoint answers `read` for any user, including one who is not a
collaborator at all, so a successful call proves nothing on its own. Do not use
`repos/{owner}/{repo}/collaborators/{user}` here — it answers `204` for `read`- and `triage`-only
collaborators, who are not the intended audience for this template.
If the result is anything else, do not use this template. Explain that it is meant for maintainers
with write access, politely ask the contributor to use one of the public templates instead, and
offer to help them file it there — their report is welcome, just not under this template. Do not
quietly switch templates while keeping the Engineering Task labels or issue type.
### Step 2: Read the Selected Template
1. Read the corresponding template file from `.github/ISSUE_TEMPLATE/` directory.
2. Identify required fields (`validations.required: true`), title prefix (`title`), labels (`labels`, if present), and issue type (`type`, if present).
For Engineering Task, read [references/engineering-task.md](references/engineering-task.md) instead — it lives outside `.github/ISSUE_TEMPLATE/` on purpose and carries its own title prefix, labels, issue type, and body structure.
### Step 3: Collect Information
Based on the selected template, ask the user for required information only. Follow the template's required fields and option constraints (for example, Platform and Priority choices).
### Step 4: Build and Preview Issue Content
Create a temp file and write the issue content:
- Use `issue_body_file="$(mktemp /tmp/gh-issue-body-XXXXXX).md"`
- Use the exact title prefix from the selected template.
- Fill content following the template body structure and section order.
- Apply labels exactly as defined by the template.
- Keep all labels when there are multiple labels.
- If template has no labels, do not add custom labels.
- Apply the issue type exactly as defined by the template's `type` field.
Preview the temp file content. **Show the file path** (e.g., `/tmp/gh-issue-body-XXXXXX.md`) and ask for confirmation before creating. **Skip this step if the user explicitly indicates no preview/confirmation is needed** (for example, automation workflows).
### Step 5: Create Issue
Use `gh issue create` command to create the issue.
Use a unique temp file for the body:
```bash
issue_body_file="$(mktemp /tmp/gh-issue-body-XXXXXX).md"
cat > "$issue_body_file" <<'EOF'
...issue body built from selected template...
EOF
```
Create the issue using values from the selected template:
```bash
gh issue create --title "<title_with_template_prefix>" --body-file "$issue_body_file"
```
If the selected template includes labels, append one `--label` per label:
```bash
gh issue create --title "<title_with_template_prefix>" --body-file "$issue_body_file" --label "<label_1_from_template>" --label "<label_2_from_template>"
```
If the selected template has no labels, do not pass `--label`.
If the selected template declares a `type`, pass it as well. `--body-file` does not carry the issue type over on its own, so omitting this leaves the issue untyped:
```bash
gh issue create --title "<title_with_template_prefix>" --body-file "$issue_body_file" --type "<type_from_template>"
```
`--template` and `--web` resolve against `.github/ISSUE_TEMPLATE/` on the default branch, so neither works for Engineering Task. Build the body locally for it.
For the other templates you may use `--template` as a starting point (use the exact template name from the repository):
```bash
gh issue create --template "<template_name>"
```
Use the `--web` flag to open the creation page in browser when complex formatting is needed:
```bash
gh issue create --web
```
Clean up the temp file after creation:
```bash
rm -f "$issue_body_file"
```
## Notes
- Must read template files under `.github/ISSUE_TEMPLATE/` (or `references/` for Engineering Task) to ensure following the correct format.
- Treat template files as the only source of truth. Do not hardcode title prefixes, labels, or issue types in this skill.
- Title must be clear and concise, avoid vague terms like "a suggestion" or "stuck从当前安装包查询 Cherry Studio 产品信息并排查运行问题。当用户询问功能、路由、快捷键、Provider、语言、Agent、频道、定时任务、Code CLI、当前版本,或报告运行错误、连接失败、配置异常并需要诊断时触发。
Develop, fix, and profile Cherry Studio in a tracked Electron instance. Use for everyday implementation, UI and interaction work, bug fixing, runtime debugging, DevTools inspection, lag or jank investigation, CPU and memory monitoring, leak checks, and startup-performance analysis; reuse a verified workspace instance across instructions and launch or replace one only when required.
当用户明确要求搜索、安装、查看、卸载或创建 Skill,或内置 Skill / 工具出现能力缺口、无法完成当前任务时触发。通过 `mcp__skills__search_skills` 搜索并用 `mcp__skills__install_skill` 安装;已安装 Skill 的查看和删除通过产品清单导航到 Skills UI;没有合适结果时调用内置 `skill-creator` 创建并验证自定义 Skill,再继续原任务。普通任务仍先尝试内置能力。
Use when Cherry Studio 用户希望报告、提交或整理 BUG、UI/UX 问题或功能建议,但未明确要求创建 GitHub Issue。
Cherry Studio first-party tool and bundled-shell routing for general agents. For straightforward local work in shell-capable sessions, run JS/TS with `bun <file>` and one-off JS tools with `bun x`; run Python with `uv run [--with <pkg>] python` and one-off Python CLIs with `uvx`; search with `rg`. Load this guide before changing project dependencies, deciding whether a tool should be ephemeral or reusable, reading or converting local Office/PDF files, coordinating or delegating across Agent Sessions, or using Cherry-owned web/browser, knowledge, persistent memory, schedules/notifications, IM channels, image generation, artifact reporting, managed CLI, skill, or MCP-server-registration capabilities—even if the user names no tool. Consult it before shell/file workarounds; live tool schemas are authoritative.
Analyze a codebase and recommend Claude Code automations (hooks, subagents, skills, plugins, MCP servers). Use when user asks for automation recommendations, wants to optimize their Claude Code setup, mentions improving Claude Code workflows, asks how to first set up Claude Code for a project, or wants to know what Claude Code features they should use.
Runs Antigravity CLI headlessly for repository analysis and coding tasks. Use when the user asks to delegate work to Antigravity CLI or compare its result with another coding agent.
Runs Claude Code non-interactively for code analysis and implementation tasks. Use when the user asks to delegate repository work to Claude Code or compare its result with another coding agent.