openclaw-agency-agents 安全扫描报告 — 低风险 | ClawSafe

15 /100

openclaw-agency-agents

管理和激活来自 agency-agents-zh 仓库的 176 个专业 AI 智能体人格

This is a legitimate agent persona management skill that clones a public GitHub repository and writes agent configuration files; no malicious behavior observed, though SKILL.md does not declare network access and setup.sh lacks version pinning.

技能名称openclaw-agency-agents

分析耗时33.4s

引擎pi

✓

可以安装

Add explicit network:READ and filesystem:WRITE declarations in SKILL.md. Pin the git clone to a specific commit hash or tag in setup.sh to prevent unexpected content changes from the upstream repository.

安全发现 3 项

严重性	安全发现	位置
低危	SKILL.md does not declare required network access 文档欺骗 The skill performs git clone and git pull operations to fetch the agency-agents-zh repository, but SKILL.md makes no mention of network access or external repository dependencies. `git clone https://github.com/jnMetaCode/agency-agents-zh` → Add a 'Required Permissions' or 'Resources' section to SKILL.md explicitly listing network:READ for the GitHub domain.	`SKILL.md:1`
低危	No version pinning on upstream repository 供应链 setup.sh clones the latest main branch with no commit hash or tag pinning. An upstream commit could introduce malicious agent personas at any time. `git clone https://github.com/jnMetaCode/agency-agents-zh.git "$REPO_DIR"` → Pin to a specific commit SHA or tag, e.g., 'git clone https://github.com/jnMetaCode/agency-agents-zh.git --depth 1 --branch v1.0.0 "$REPO_DIR"'	`scripts/setup.sh:53`
提示	Script header name mismatch 文档欺骗 setup.sh header comments refer to 'openmaic-agents-lite' while SKILL.md describes 'openclaw-agency-agents'. Minor inconsistency. `# setup.sh - 初始化 openmaic-agents-lite skill` → Update script header to match the SKILL.md name 'openclaw-agency-agents'.	`scripts/setup.sh:1`

资源类型	声明权限	推断权限	状态	证据
文件系统	`NONE`	`WRITE`	✓ 一致	setup.sh:53, activate.sh:93 writes files to $WORKSPACE_DIR
网络访问	`NONE`	`READ`	✓ 一致	setup.sh:53 'git clone https://github.com/jnMetaCode/agency-agents-zh.git', upda…
命令执行	`NONE`	`WRITE`	✓ 一致	All scripts execute bash; activate.sh creates config files via heredoc and file …

目录结构

7 文件 · 17.3 KB · 576 行

Shell 6f · 458L Markdown 1f · 118L

├─ ▾ 📁 scripts

│ ├─ 🔧 activate.sh Shell 153L · 4.5 KB

│ ├─ 🔧 list.sh Shell 77L · 2.4 KB

│ ├─ 🔧 restore.sh Shell 65L · 1.4 KB

│ ├─ 🔧 search.sh Shell 69L · 2.0 KB

│ ├─ 🔧 setup.sh Shell 59L · 2.2 KB

│ └─ 🔧 update.sh Shell 35L · 765 B

└─ 📝 SKILL.md Markdown 118L · 4.0 KB

依赖分析 1 项

包名	版本	来源	已知漏洞	备注
`agency-agents-zh`	`latest (unpinned)`	https://github.com/jnMetaCode/agency-agents-zh.git	否	Clones latest main branch with no commit/tag pinning

安全亮点

✓ No credential harvesting or environment variable enumeration observed

✓ No base64, obfuscation, or anti-analysis patterns detected

✓ No network exfiltration or C2 communication patterns

✓ No sensitive path access (~/.ssh, ~/.aws, .env)

✓ No reverse shell, RCE, or arbitrary command injection

✓ All operations are scoped to a specific skill directory and workspace directory

✓ Backup mechanism provides config restore safety

✓ Uses set -euo pipefail for safe shell scripting