技能 m3-expressive 审计历史
📦

审计历史

m3-expressive - 1 审计

2026年10月10日 19:38

Of 400 supplied alerts, 396 are false positives; four confirm conditional filesystem and network risks in the optional image refresh script. No evidence found of malicious intent, credential theft, or prompt injection in the reviewed context. The analyzer withheld 21 additional matches, which require manual review before automatic publication. Static review was capped at 400/421 representative findings; omitted static matches are unconfirmed, so automatic publishing stays disabled until manual review.

26
已扫描文件
12,538
分析行数
8
复核项
0
已忽略误报
能力复核项 (4)

这些是真实的本地能力,对此技能可能是预期行为,因此需要复核,但不会按已确认恶意行为计分。

中
Node.js fs operations
for (const s of new Set(found.map(f => f.rel.split('/')[0]))) fs.mkdirSync(path.join(outdir, s), { r
Image paths use remote route slugs with separators replaced, but allow ".." as a directory name. Malformed upstream data can cause writes outside outdir.
中
Node.js fs operations
fs.writeFileSync(dest, buf);
Image paths use remote route slugs with separators replaced, but allow ".." as a directory name. Malformed upstream data can cause writes outside outdir.
中
Synchronous file operations
fs.writeFileSync(dest, buf);
Image paths use remote route slugs with separators replaced, but allow ".." as a directory name. Malformed upstream data can cause writes outside outdir.
中
HTTP/HTTPS get
https.get(url, { headers: { 'User-Agent': 'Mozilla/5.0' } }, res => {
The image downloader accepts upstream image URLs and follows redirects without host or private-address restrictions. Manipulated upstream content could trigger requests to internal HTTPS services.

风险因素

📁 文件系统访问 (24)
⚡ 包含脚本 (7)
⚙️ 外部命令 (50)
🌐 网络访问 (6)
审计者: codex