Reviews code changes for correctness, security, and maintenance risks to provide actionable defect reports.
- GitHub stars
- 72
- Files
- 13
- Last commit
- Nov 10, 2026
- License
- MIT
- Allowed tools
- —
What does the lov-code-review skill do?
lov-code-review reviews code changes for correctness, security, and maintenance risks to provide actionable defect reports.
Summary generated from the skill's SKILL.md.
About the lov-code-review skill
Developers looking to identify defects in their code changes can use this tool to analyze diffs for correctness, security, and maintenance risks. It functions by reading specific code differences, relevant calls, and project rules to provide actionable reports that include file line numbers, trigger conditions, and potential consequences. The process focuses on finding reproducible defects, input boundaries, permission issues, secrets, and performance regressions rather than merely flagging stylistic preferences.
An agent should load this skill when a user requests a review of a code change or asks to find actionable defects. It relies on the current host's file access, search, CLI, and browser capabilities to understand the context. The tool operates in a read-only mode during the analysis phase to protect uncommitted content, and it only performs modifications or commits if the user explicitly requests a fix. It integrates with the lov-branding-consistency skill and uses Python 3.8+ and PyYAML for profile management and validation.
What an agent can do with it
- 1Identify security vulnerabilities in diffsAsk the agent to check a specific code change for secrets, permission issues, or security risks.
- 2Generate actionable defect reportsRequest a detailed report that includes line numbers and fix suggestions for identified bugs.
- 3Perform read-only code analysisInstruct the agent to review changes without making any actual modifications to the codebase.
Written from the skill's SKILL.md. Every use case is checked against a quote from it.
How to install the lov-code-review skill
Each command copies the skill folder into the client's skills directory. Restart the client or start a new session to load it.Install the lov-code-review skill in Plugin
/plugin marketplace add lovstudio/skills /plugin install dev-tools@lovstudio
Install the lov-code-review skill in Claude Code
~/.claude/skills/lov-code-review/SKILL.md
$ npx degit lovstudio/skills/skills/code-review ~/.claude/skills/lov-code-reviewInstall the lov-code-review skill in Claude Code (project)
.claude/skills/lov-code-review/SKILL.md
$ npx degit lovstudio/skills/skills/code-review .claude/skills/lov-code-reviewInstall the lov-code-review skill in Codex
~/.agents/skills/lov-code-review/SKILL.md
$ npx degit lovstudio/skills/skills/code-review ~/.agents/skills/lov-code-reviewInstall the lov-code-review skill in Gemini CLI
~/.gemini/skills/lov-code-review/SKILL.md
$ npx degit lovstudio/skills/skills/code-review ~/.gemini/skills/lov-code-reviewInstall the lov-code-review skill in Claude.ai
npx degit lovstudio/skills/skills/code-review lov-code-review && zip -r lov-code-review.zip lov-code-review # claude.ai → Customize → Skills → + → Create skill → Upload a skill → lov-code-review.zip
Automated checks
Ran Nov 15, 2026Full report →Passed our automated checks. That is not a security audit: read the scripts before you let an agent run them. How checks work · What to fix for more points
- Source repository linked and reachablelovstudio/skills/skills/code-review is public and reachablegithub.com/lovstudio/skills/tree/HEAD/skills/code-review ↗Pass
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit 2026-11-10github.com/lovstudio/skills/commits ↗Pass
- No download-and-run or credential theft in scripts2 scripts read; no download-and-run, remote eval, base64 execution or credential uploadPass
- No hidden instructions in SKILL.mdNo override phrases, safeguard switches, credential requests or hidden charactersgithub.com/lovstudio/skills/blob/HEAD/skills/code-review/SKI ↗Pass
- Pre-approved tools are scopedNo pre-approved tools; the agent asks before using onegithub.com/lovstudio/skills/blob/HEAD/skills/code-review/SKI ↗Pass
- SKILL.md follows the Agent Skills formatname "lov-code-review" differs from the folder "code-review"github.com/lovstudio/skills/blob/HEAD/skills/code-review/SKI ↗Finding
SKILL.md
View on GitHub代码审阅官 · Code Reviewer
审查实际差异中的正确性、安全和维护风险并给出可定位意见。
Triggers
Activate when
- “审查实际差异中的正确性、安全和维护风险并给出可定位意见。”
- “Review a code change for actionable defects.”
Do not activate when
- 只是查询本 Skill 的说明,或请求与上述结果无关的任务;不执行实际业务操作。
- 用户仅要预览或审查时,不进入修改、提交或发布分支。
Execution boundary
自然语言请求即可触发;无需旧 slash 路径、参数插值或指定助手。明确解析当前请求中的 项目、目标文件、选项与输出位置;用当前宿主实际提供的文件、搜索、CLI 和浏览器能力。 项目依赖版本与外部 API 在执行时核实,不能假设示例是现行配置。随包脚本从 Skill 根解析, 业务文件从目标项目根解析。先读当前状态,保护已有未提交内容与其他任务的暂存区。 分析、预览请求保持只读;修改、提交、推送、部署和发布各依当前请求的明确范围执行。 不绕过保护、自动发送消息、强制结束用户进程或抢前台。失败保留可诊断原始错误。
Workflow
-
读取明确范围的 diff、相关调用与项目规则;无范围时检查当前变更,不假定审查整个仓库。
-
优先检查可复现缺陷、输入边界、权限、秘密、错误处理、状态、副作用及性能退化,结合调用上下文判断。
-
区分真正影响行为的问题和风格偏好,不把通用检查清单当成已发现缺陷;必要时运行有针对性的只读验证。
-
按严重度给具体文件行号、触发条件、后果和修复建议;没有可行动问题时如实说明,不凑发现。
-
审查默认不改代码、不提交、不发送 review 评论;用户要求修复时再按范围实施。
Composition
执行前读取 能力组合,按明确制品交接相邻能力。
Runtime context (shared)
运行前读取本包 skill.yaml 与 Profile 合同。优先级为当前请求、
项目上下文、本 Skill records、共享 preferences、brand/user Profile、安全默认值。
只读取声明字段;没有专用运行时的宿主可使用 scripts/profile_store.py 读取共享 Profile。
配置缺失只问影响结果的一个问题。用户明确要求长期保存的值通过该脚本原子写入,
报告实际路径;不保存推断、凭据或其他任务的资料。
Files in this skill
13lov-code-review skill questions
How do I install the lov-code-review skill in Claude Code?
Run npx degit lovstudio/skills/skills/code-review ~/.claude/skills/lov-code-review. Claude Code loads skills from ~/.claude/skills (all projects) or .claude/skills (one project) and uses a skill when its description matches the task.
Which clients can use the lov-code-review skill?
Any client that reads the Agent Skills format: Claude Code, Codex, Gemini CLI and claude.ai (upload as a ZIP). Scripts in the skill run only where the client allows code execution.
What does this skill review?
The skill reviews actual code differences for correctness, security, and maintenance risks, providing actionable feedback.
Does the skill automatically modify or commit code?
No, the review process is read-only by default and does not modify, commit, or send review comments unless specifically requested by the user.
How are defects reported?
Defects are reported by severity, including specific file line numbers, trigger conditions, consequences, and suggestions for fixing.