📦

审计历史

schema-visualizer - 8 审计

版本对比

各审计版本间的能力与发现变化,最新在前。

版本日期结果复核项相对上一版变化
v8 最新2026年7月5日 07:31 无已确认发现0能力无变化
v7 2026年7月5日 07:31 无已确认发现0能力无变化
v6 2026年6月29日 09:14 1 已确认0能力无变化
v5 2026年1月16日 23:31 无已确认发现0能力无变化
v4 2026年1月16日 23:31 无已确认发现0外部命令
v3 2026年1月10日 12:57 无已确认发现0能力无变化
v2 2026年1月10日 12:57 无已确认发现0能力无变化
v1 2026年1月10日 12:57 无已确认发现0基线

2026年7月5日 07:31

All static findings are false positives caused by Markdown fences, inline code formatting, and database schema keywords in SKILL.md. No prompt injection, data exfiltration intent, or bundled executable code was found.

1
已扫描文件
846
分析行数
1
复核项
0
已忽略误报
审计者: codex

2026年7月5日 07:31

All static findings are false positives caused by Markdown fences, inline code formatting, and database schema keywords in SKILL.md. No prompt injection, data exfiltration intent, or bundled executable code was found.

1
已扫描文件
846
分析行数
1
复核项
0
已忽略误报
审计者: codex

2026年6月29日 09:14

Static analysis reported many command-execution and weak-cryptography alerts, but most are false positives from Markdown code fences, inline identifiers, and schema field names. The skill is legitimate documentation guidance, but it includes database dump and ORM commands that can access sensitive schema metadata, so publication should include a medium-risk warning.

1
已扫描文件
846
分析行数
2
复核项
3
已忽略误报

已确认安全问题 (1)

中
Database Schema Extraction Commands Require Review
The skill documents shell commands for pg_dump, mysqldump, SchemaSpy, Prisma, TypeORM, Django, and Rails ERD generation. These commands are appropriate for schema visualization, but they may read database metadata or expose credentials if users run them against sensitive systems without review.
The referenced lines contain real external command examples. The commands support the skill purpose, so the concern is controlled operational risk rather than malicious behavior.
已忽略静态误报 (3)

这些静态命中已被语义复核判定为误报,或只命中了 schema 定义里的词;这里保留展示用于透明度,但不影响质量评分。

低
Backtick Execution Alerts Are Markdown Syntax
The static Ruby backtick execution alerts occur in Markdown code fences and inline identifiers used to show examples. I found no Ruby code execution path or instruction to execute user-controlled shell text.
The cited locations are fenced Markdown examples, not executable Ruby source. The surrounding file is documentation-only skill text.
低
Weak Cryptography Alerts Are Substring Matches
The high-severity weak cryptography alerts appear to match words such as description and password_hash in schema examples. I found no MD5, SHA-1, DES, RC4, or cryptographic implementation guidance.
The inspected lines are ordinary schema documentation fields. The only password-related text describes a bcrypt hashed password column, which is not weak cryptography.
低
Reconnaissance Alerts Reflect Intended Schema Inspection
The scanner flagged many schema inspection references as reconnaissance. The skill is designed to inspect tables, columns, indexes, constraints, ORM models, and migrations for documentation, with no evidence of covert collection or exfiltration.
The behavior is explicitly aligned with the advertised schema visualization purpose. No network destination, credential harvesting, or stealth behavior was found.

风险因素

检测到的模式

External Database and ORM Tool Commands
审计者: codex

2026年1月16日 23:31

All 187 static findings are FALSE POSITIVES. This is a pure documentation skill containing only markdown text and example code snippets. The scanner incorrectly flagged documentation patterns (bcrypt mentions, command examples, database terms) as security issues. No executable code, network access, filesystem access, or external command execution exists. The pre-computed report correctly assessed this as safe.

2
已扫描文件
1,024
分析行数
1
复核项
0
已忽略误报
审计者: claude

2026年1月16日 23:31

All 187 static findings are FALSE POSITIVES. This is a pure documentation skill containing only markdown text and example code snippets. The scanner incorrectly flagged documentation patterns (bcrypt mentions, command examples, database terms) as security issues. No executable code, network access, filesystem access, or external command execution exists. The pre-computed report correctly assessed this as safe.

2
已扫描文件
1,024
分析行数
1
复核项
0
已忽略误报
审计者: claude

2026年1月10日 12:57

Pure prompt-based skill with no executable code, no network access, no filesystem access, and no external command execution. Contains only AI instructions for database schema visualization.

1
已扫描文件
846
分析行数
0
复核项
0
已忽略误报
此次已完成的审计未记录已确认的安全发现。
审计者: claude

2026年1月10日 12:57

Pure prompt-based skill with no executable code, no network access, no filesystem access, and no external command execution. Contains only AI instructions for database schema visualization.

1
已扫描文件
846
分析行数
0
复核项
0
已忽略误报
此次已完成的审计未记录已确认的安全发现。
审计者: claude

2026年1月10日 12:57

Pure prompt-based skill with no executable code, no network access, no filesystem access, and no external command execution. Contains only AI instructions for database schema visualization.

1
已扫描文件
846
分析行数
0
复核项
0
已忽略误报
此次已完成的审计未记录已确认的安全发现。
审计者: claude