Token导航 LogoToken导航TokenDH.com
开发需要联网github未标认证来源可访问许可证需确认审计提醒

git-intelligence吉特智能

Agent Skill

git-intelligence 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

214

周安装

9

GitHub Stars

4

下载量

75
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:git-intelligence(吉特智能)
来源仓库:https://github.com/wghust/stark-skills
仓库路径:skills/git-intelligence
安装命令:
npx skills add https://github.com/wghust/stark-skills --skill git-intelligence
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/wghust/stark-skills --skill git-intelligence

简介

git-intelligence 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息。

  • 适合在 Codex、Claude、Cursor、Gemini CLI 中围绕仓库状态、代码变更或协作事项进行整理。
  • 通过 npx skills add 命令从指定 GitHub 仓库安装使用。
  • 安装前需确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写操作。
  • 适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

SKILL.md

git-intelligence Skill

Zero-install git analysis skill. The agent runs read-only git commands directly and uses its own reasoning to produce structured reports — no CLI tools, no API keys, no compilation.

When to Use

Activate when the user asks to:

  • "Summarize commits from the last N days / this week"
  • "Generate release notes between two tags"
  • "Analyze this PR / branch"
  • "Is this branch / PR risky?"
  • "What does this branch affect?"
  • "Who should review this PR?"

Step 1 — Verify Git Repository

Before any analysis, run:

git rev-parse --git-dir
  • If it returns a path → proceed
  • If it fails → stop and tell the user: "This directory is not a git repository. Please navigate to your project root first."

Step 2 — Intent Mapping

User saysCapabilityGo to
"Summarize commits" / "What changed this week?"summarize-commitsStep 3A
"Release notes" / "What's in v1.3?"release-notesStep 3B
"Analyze this PR" / "Full PR review"analyze-prStep 3C
"Is this risky?" / "Risk check"detect-riskStep 3D
"What does this branch affect?" / "Impact?"change-impactStep 3E

Step 3A — Summarize Commits

Collect data:

# Convert duration: 7d=7 days ago, 2w=14 days, 1mo=1 month
git log --since="<ISO date>" --oneline --no-merges

Analyze: Classify each commit message by keyword:

  • feat / add / new / implement / supportFeatures
  • fix / bug / patch / resolve / hotfixFixes
  • refactor / clean / move / renameRefactor
  • perf / optim / speed / cachePerformance
  • BREAKING CHANGE or !: notation → Breaking Changes
  • Unmatched → Features (default)

Use your own understanding to improve on keyword matching when the meaning is clear.

Output format:

## Commit Summary (since <duration>)

### Features
- <commit message>

### Fixes
- <commit message>

### Performance
- <commit message>

Omit empty sections. If no commits found: "No commits found since."


Step 3B — Release Notes

Collect data:

# Verify tags exist first
git rev-parse --verify refs/tags/<from>
git rev-parse --verify refs/tags/<to>

# Get commits between tags
git log <from>..<to> --oneline --no-merges

If a tag doesn't exist, stop and report: "Tag <tag> not found. Run git tag -l to list available tags."

Analyze: Same classification as Step 3A, but rewrite commit messages into user-facing language where appropriate.

Output format:

## Release Notes: <from> → <to>

### Breaking Changes
- <change>

### Features
- <feature description>

### Fixes
- <fix description>

### Performance
- <improvement>

Omit empty sections (especially Breaking Changes).


Step 3C — Analyze PR

Combines risk detection + impact analysis + reviewer suggestion.

Collect data:

# Auto-detect base branch
git symbolic-ref refs/remotes/origin/HEAD   # e.g. refs/remotes/origin/main → "main"
# If fails → use "master"

# Get diff statistics
git diff <base>...<branch> --numstat

# For reviewer suggestion: blame top changed files (up to 5)
git blame --porcelain -- <file>

Analyze:

  1. Compute risk score (same rules as Step 3D)
  2. Map changed files → logical modules (group by top 2 path segments)
  3. Parse blame output → count lines per author email → take top 3 authors by blame-line count
  4. If fewer than 2 blame authors found, add role fallbacks: auth/ → "security team", payment/ → "payments team", infra/ → "infra team"
  5. Suggest test cases based on affected modules and sensitive areas

Output format:

## PR Analysis: <branch>

### Summary
<1–2 sentence description of what the branch changes>

### Impacted Modules
- <module>

### Risk Score
<🔴/🟡/🟢> <score>/10

### Suggested Reviewers
- <name or role>

### Suggested Tests
- <test scenario>

Step 3D — Detect Risk

Collect data:

git symbolic-ref refs/remotes/origin/HEAD   # base branch, fallback to master
git diff <base>...<branch> --numstat

Risk scoring (heuristic):

ConditionPoints
Files changed > 50+3
Total lines changed > 1000+3
Sensitive folder touched: auth/, payment/, infra/, middleware/, cache/+2 per folder (max +4)
Test files < 10% of changed files+1

Start at 1. Clamp final score to [1, 10]. Add your own reasoning on top of the heuristic if the code context warrants it.

Risk label: 🔴 score ≥ 8 · 🟡 score 5–7 · 🟢 score ≤ 4

Output format:

## Risk Assessment: <branch>

**Risk Score**: 🔴 8/10

**Reasons**:
- Large diff (1200 lines)
- Middleware module modified
- Low test coverage for changed files

Step 3E — Change Impact

Collect data:

git symbolic-ref refs/remotes/origin/HEAD   # base branch, fallback to master
git diff <base>...<branch> --numstat

Analyze:

  • Group files by top 2 path segments → logical module names
  • Identify sensitive modules (auth, payment, infra, middleware, cache) and flag them
  • Use your understanding of the codebase to infer downstream effects

Output format:

## Change Impact: <branch>

### Modules Affected
- <module path> — <brief impact note>

### Potential Risks
- <risk description>

Handling Edge Cases

SituationAction
No commits in range"No commits found for this period."
Tag not foundReport tag name, suggest git tag -l
Branch doesn't existReport error from git diff, suggest git branch -a
Large diff (> 100 files)Analyze first 100 files, note truncation
Empty repository"No commits yet in this repository."
Detached HEADUse git rev-parse HEAD as current ref

Read-Only Guarantee

This skill ONLY runs read-only git commands: git log, git diff, git blame, git rev-parse, git symbolic-ref, git tag, git show.

It NEVER runs: git commit, git checkout, git merge, git push, git stash, git reset, git add.

No files are created or modified in the repository.

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

35%
按下载量换算26

Claude

31.15%
按下载量换算23

Cursor

16.79%
按下载量换算13

Gemini CLI

9.56%
按下载量换算7

安全审计

Gen Agent Trust Hub

可疑

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills