code-review

星标8

分支2

更新时间2026年2月16日 10:39

Review code changes for correctness, security, test coverage, and style.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

来源

bis-code

bis-code/claude-toolkit

下载

/code-review

Spawns the code-reviewer agent to analyze your current changes and provide structured feedback.

Gather context — detect the scope of changes to review:
- If arguments are provided (file paths or PR number), use those as scope
- Otherwise, use git diff (unstaged) and git diff --cached (staged) to determine what changed
- Run git log --oneline -5 for recent commit context
Spawn the agent — use the Task tool:
```
Task tool with subagent_type="code-reviewer"
```
Pass in the prompt:
- The diff output (or file contents if reviewing specific files)
- The list of changed files
- Any project conventions from .claude/rules/ that apply
Present findings — relay the agent's review to the user:
- Group by severity: CRITICAL > WARNING > NIT > PRAISE
- Include file paths and line numbers
- Show suggested fixes for each finding
Offer follow-up actions:
- "Fix critical issues?" — apply suggested fixes
- "Run tests?" — verify fixes do not break anything
- "Review again?" — re-run after changes

同仓库更多 Skills

同仓库

Interactive agent builder: guides the user through defining a new agent as a behavioral spec, then generates the agent .md and installs it.

Cross-repo investigation: research planning issues, break down into sub-issues, generate ralph session prompts per repo.

Three-phase QA orchestrator: parallel scan, interactive triage, guided fix.

Interactive feature builder with subagent delegation and approval gates.

Interactive rule builder: guides the user through defining a new rule as a constraint doc, then generates the rule .md and installs it.

Interactive skill builder: guides the user through defining a new skill as a structured SOP, then generates the SKILL.md and installs it.

name	code-review
description	Review code changes for correctness, security, test coverage, and style.

Spawns the code-reviewer agent to analyze your current changes and provide structured feedback.

Gather context — detect the scope of changes to review:
- If arguments are provided (file paths or PR number), use those as scope
- Otherwise, use git diff (unstaged) and git diff --cached (staged) to determine what changed
- Run git log --oneline -5 for recent commit context
Spawn the agent — use the Task tool:
```
Task tool with subagent_type="code-reviewer"
```
Pass in the prompt:
- The diff output (or file contents if reviewing specific files)
- The list of changed files
- Any project conventions from .claude/rules/ that apply
Present findings — relay the agent's review to the user:
- Group by severity: CRITICAL > WARNING > NIT > PRAISE
- Include file paths and line numbers
- Show suggested fixes for each finding
Offer follow-up actions:
- "Fix critical issues?" — apply suggested fixes
- "Run tests?" — verify fixes do not break anything
- "Review again?" — re-run after changes