changelog-from-commits
关于
This skill generates user-facing changelog entries from git commit history, ideal for preparing release notes or summarizing changes since a previous version. It transforms raw git log output into formatted summaries suitable for end users, not contributors. Use it when cutting releases or when asked "what changed since last version" to create clean, readable change documentation.
快速安装
Claude Code
推荐npx skills add rockscy/solo-skills -a claude-code/plugin add https://github.com/rockscy/solo-skillsgit clone https://github.com/rockscy/solo-skills.git ~/.claude/skills/changelog-from-commits在 Claude Code 中复制并粘贴此命令以安装该技能
技能文档
Changelog from Commits / 提交转更新日志
When to use
- The user is cutting a release and needs user-facing release notes.
- There is a clear range: a tag, a date, or a commit SHA to diff from.
- The audience is users, not contributors (Conventional Commits already exists for the latter).
When NOT to use
- The repo has no user-facing surface (internal libraries, scripts) — a commit log is fine.
- The range covers months of work. Break it into smaller releases first.
- The user wants a marketing post — that's launch-tweet, not a changelog.
Approach
- Get the raw commit list:
git log <prev-tag>..HEAD --pretty=format:"%h %s%n%b" --no-merges - Group commits into three buckets, in this order:
- Added — net-new user-visible capability.
- Changed / Fixed — behavior that users would notice.
- Internal — refactors, deps, docs (collapse into a single line at the bottom).
- Translate from dev-speak to user-speak.
- ❌ "refactor: extract validation middleware"
- ✅ (skip — internal)
- ❌ "fix: handle null in payment.processor"
- ✅ "Fixed a crash when applying coupons to gift cards."
- Drop noise. Typo fixes, README tweaks, CI tweaks, version bumps → cut.
- Each bullet is one sentence, past tense, user perspective.
- If a bullet needs more than one sentence, it deserves its own blog post — link out.
Output format
## v<version> — YYYY-MM-DD
### Added
- <bullet>
### Changed
- <bullet>
### Fixed
- <bullet>
### Internal
- <one-line summary>: deps bumped, tests added, docs polished.
Skip empty sections entirely.
Worked example
Input commits:
a1b2c3d feat: add CSV export to reports
e4f5g6h fix: correct rounding in monthly totals
1a2b3c4 chore: bump axios to 1.7.5
9z8y7x6 refactor: split UserService
5d4c3b2 docs: fix typo in README
abcdef0 feat: dark mode toggle
Output:
## v1.4.0 — 2026-04-30
### Added
- CSV export for any report — find it in the report toolbar.
- Dark mode toggle in Settings → Appearance.
### Fixed
- Monthly totals now round consistently with daily totals (off-by-1¢ on edge cases).
### Internal
- Dependency updates and a service refactor; no user impact expected.
Hard rules
- Past tense. "Added X", not "Adds X".
- One bullet, one sentence. If you need a second sentence, you're explaining implementation.
- Link to docs/blog if a feature needs context. Don't inline tutorials.
中文版
何时使用
- 用户在发布版本,需要面向用户的更新说明。
- 有明确的范围(上个 tag、某个日期、某个 SHA)。
- 受众是用户,不是贡献者(Conventional Commits 给后者用)。
何时不使用
- 仓库无用户访问面(内部库 / 脚本),commit 日志够了。
- 范围跨度几个月——先拆成小版本。
- 用户其实想要营销稿——那是 launch-tweet 的活。
步骤
- 拿提交列表:
git log <prev-tag>..HEAD --pretty=format:"%h %s%n%b" --no-merges - 分三组:
- 新增(Added)——新功能。
- 变更/修复(Changed/Fixed)——用户能感知的行为变化。
- 内部(Internal)——重构、依赖、文档——折叠成一行。
- 翻译成用户能懂的话。
- 过滤噪声:错别字、CI 调整、版本号 bump → 删掉。
- 每条一句话,过去时,从用户视角写。
硬规则
- 过去时。
- 一条一句。需要两句说明你在解释实现细节,删掉。
- 复杂功能链接到博客或文档,不要在 changelog 里写教程。
GitHub 仓库
相关推荐技能
content-collections
元Content Collections 是一个 TypeScript 优先的构建工具,可将本地 Markdown/MDX 文件转换为类型安全的数据集合。它专为构建博客、文档站和内容密集型 Vite+React 应用而设计,提供基于 Zod 的自动模式验证。该工具涵盖从 Vite 插件配置、MDX 编译到生产环境部署的完整工作流。
polymarket
元这个Claude Skill为开发者提供完整的Polymarket预测市场开发支持,涵盖API调用、交易执行和市场数据分析。关键特性包括实时WebSocket数据流,可监控实时交易、订单和市场动态。开发者可用它构建预测市场应用、实施交易策略并集成实时市场预测功能。
creating-opencode-plugins
元该Skill帮助开发者创建OpenCode插件,用于接入命令、文件、LSP等25+种事件。它提供了插件结构、事件API规范和JavaScript/TypeScript实现模式,适合需要拦截操作、扩展功能或自定义事件处理的场景。开发者可通过它快速构建响应式模块来增强OpenCode AI助手的能力。
sglang
元SGLang是一个专为LLM设计的高性能推理框架,特别适用于需要结构化输出的场景。它通过RadixAttention前缀缓存技术,在处理JSON、正则表达式、工具调用等具有重复前缀的复杂工作流时,能实现极速生成。如果你正在构建智能体或多轮对话系统,并追求远超vLLM的推理性能,SGLang是理想选择。
