ai-code-review

分类: 工具与效率 | 上传者: transilienceai | 下载: 0 | 版本: v1.0（最新）

针对 AI 安全问题进行深度代码扫描 — 提示注入、提示中的 PII、硬编码密钥、无人看守的代理。

更新日志: Source: GitHub https://github.com/transilienceai/shasta

目录结构

当前层级: tree/main/.claude/skills/ai-code-review/

📄 SKILL.md 1.8 KB

SKILL.md

---
name: ai-code-review
description: Deep code scan for AI security issues — prompt injection, PII in prompts, hardcoded keys, unguarded agents.
user-invocable: true
---

# AI Code Review

You are performing a deep AI security review of a code repository for a founder. Focus on practical, actionable findings.

## What to do

Read `shasta.config.json` for `python_cmd`. Scan either the current directory or a specified GitHub repo.

### Run code review

```bash
<PYTHON_CMD> -c "
import json
from pathlib import Path
from whitney.code.scanner import scan_repository

# Scan current project directory
findings = scan_repository(Path('.'))

# Group by severity
from collections import defaultdict
by_severity = defaultdict(list)
for f in findings:
    by_severity[f.severity.value].append({
        'check_id': f.check_id,
        'title': f.title,
        'file': f.details.get('file_path', 'unknown'),
        'line': f.details.get('line_number', '?'),
        'snippet': f.details.get('code_snippet', ''),
        'remediation': f.remediation,
    })

print(json.dumps({
    'total': len(findings),
    'critical': len(by_severity.get('critical', [])),
    'high': len(by_severity.get('high', [])),
    'medium': len(by_severity.get('medium', [])),
    'low': len(by_severity.get('low', [])),
    'findings': dict(by_severity),
}, indent=2))
"
```

### Present results

For each finding:
- Show the file path and line number
- Show the code snippet (3 lines of context)
- Explain what the risk is in plain English
- Provide specific remediation steps

Group by severity: CRITICAL (fix now) → HIGH (fix this sprint) → MEDIUM (fix this month) → LOW (track)

### Tone
- Be specific about what's wrong and how to fix it
- Show the actual code that's problematic
- Provide the fixed code where possible

登录后下载/点赞/收藏 ❤ 97 | ★ 0

ai-code-review

目录结构

SKILL.md

举报内容

提示