yaoyao-memory 安全扫描报告 — 低风险 | ClawSafe

15 /100

yaoyao-memory

四层渐进式长时记忆系统，让 AI 跨会话保持上下文、沉淀知识、持续进化

Legitimate memory management skill with straightforward file operations and declared IMA sync capability. Minor documentation issues but no malicious behavior.

技能名称yaoyao-memory

分析耗时44.2s

引擎pi

✓

可以安装

Safe to use. Consider clarifying the IMA credential requirements and the rm command in documentation for better user understanding.

安全发现 3 项

严重性	安全发现	位置
低危	IMA credential access not formally declared 文档欺骗 The skill accesses IMA credentials from environment variables and ~/.config/ima/ files but this is only explained in usage text, not formally declared as a capability. `client_id = os.environ.get('IMA_OPENAPI_CLIENTID')` → Add environment:READ to declared capabilities if this access is considered a resource usage.	`scripts/sync_ima.py:13`
低危	Dangerous rm commands in documentation 文档欺骗 SKILL.md line 887 shows 'rm -rf ~' style commands in the uninstall section. This is for user guidance but uses dangerous patterns. `rm -rf ~/.openclaw/workspace/MEMORY.md` → Use safer deletion patterns or clarify these are user commands, not skill execution.	`SKILL.md:887`
提示	IMA credentials stored in plain text files 敏感访问 The documentation instructs users to store API credentials in plain text files at ~/.config/ima/. While this is a common pattern, it could be improved with a warning about file permissions. `echo 'your_client_id' > ~/.config/ima/client_id` → Add note about setting file permissions (chmod 600) for credential files.	`SKILL.md:45`

资源类型	声明权限	推断权限	状态	证据
文件系统	`WRITE`	`WRITE`	✓ 一致	All scripts use pathlib for file operations
网络访问	`NONE`	`READ`	✓ 一致	sync_ima.py:52 makes HTTP POST to ima.qq.com
环境变量	`NONE`	`READ`	✓ 一致	sync_ima.py:13-19 reads IMA_OPENAPI_* env vars
命令执行	`NONE`	`NONE`	—	No subprocess or shell commands in code

1 严重 3 项发现

💀

严重危险命令危险 Shell 命令

rm -rf ~

SKILL.md:887

🔗

中危外部 URL 外部 URL

https://ima.qq.com/agent-interface

SKILL.md:31

🔗

中危外部 URL 外部 URL

https://ima.qq.com/

scripts/sync_ima.py:40

目录结构

10 文件 · 78.4 KB · 2711 行

Python 7f · 1517L Markdown 3f · 1194L

├─ ▾ 📁 references

│ ├─ 📝 heartbeat-maintenance.md Markdown 125L · 2.7 KB

│ └─ 📝 memory-lifecycle.md Markdown 113L · 2.4 KB

├─ ▾ 📁 scripts

│ ├─ 🐍 cleanup.py Python 142L · 4.5 KB

│ ├─ 🐍 generate_index.py Python 196L · 5.9 KB

│ ├─ 🐍 init_memory.py Python 379L · 11.0 KB

│ ├─ 🐍 migrate.py Python 129L · 4.1 KB

│ ├─ 🐍 promote.py Python 224L · 6.9 KB

│ ├─ 🐍 summarize.py Python 219L · 6.6 KB

│ └─ 🐍 sync_ima.py Python 228L · 7.6 KB

└─ 📝 SKILL.md Markdown 956L · 26.6 KB

安全亮点

✓ No subprocess or shell command execution in code

✓ No base64 encoding, obfuscation, or anti-analysis techniques

✓ No credential exfiltration - IMA keys stay local for API authentication only

✓ Clean Python standard library usage (pathlib, json, urllib)

✓ All file operations are within documented workspace scope

✓ No network calls to unknown/undeclared external IPs

✓ Supports --dry-run for safe preview of destructive operations

✓ No hidden functionality - all scripts are straightforward file/text processing