secureclaw

Security hardening toolkit for OpenClaw. Run audits, apply fixes, scan skills, monitor costs and memory integrity.

执行命令联网改身份文件写文件读文件读环境变量（配置）严重 98 · 高危 134adversa-ai/secureclaw

它会碰到什么

扫了多少66 个文本文件，963 KB

它会碰到什么执行命令联网改身份文件写文件读文件读环境变量（配置）

命中总数479 处

命中统计严重 98 · 高 134 · 中 205 · 低 42

关于「读环境变量（配置）」：这个技能会读 process.env 之类的环境变量，但读到的都是端口、目录、超时这类配置项，没有读取密钥类变量。扫描规则原本把「读环境变量」一律算作「读凭据」，本站按变量名做了细化区分，命中明细仍如实列在下面。

逐条看命中（30 条严重或高危）

严重 demo.ts:74cred-paths
```
// Plaintext API keys in .env
```
严重 demo.ts:76cred-paths
```
path.join(DEMO_DIR, '.env'),
```
严重 demo.ts:140cred-paths
```
fs.readFileSync("~/.openclaw/.env");
```

严重 demo.ts:155cred-paths

console.log('  Created .env with plaintext API keys (permissions 644)');

严重 README.md:66cred-paths

AI agents with access to your files, credentials, email, and the internet are a fundamentally different security surface than traditional software. An agent tha

严重 README.md:70cred-paths

**Layer 1 -- Audit.** 56 automated checks across 8 categories scan your OpenClaw installation for known misconfigurations: exposed gateway ports, weak file perm

严重 README.md:246cred-paths

| File permissions | `.env` file (needs 600), installation directory (needs 700) |

严重 README.md:247cred-paths

| Credential exposure | Scans for API key patterns (`sk-ant-`, `sk-proj-`, `xoxb-`, `ghp_`, `AKIA`) outside `.env` |

严重 README.md:251exec-pipe-to-shell

| Supply chain | Scans installed skills for dangerous patterns (`curl\|sh`, `eval()`, `webhook.site`) |

严重 README.md:326cred-paths

| `.env` permissions | Sets `.env` to mode 600 (owner read/write only). |

严重 README.md:326cred-paths

| `.env` permissions | Sets `.env` to mode 600 (owner read/write only). |

严重 README.md:441exec-pipe-to-shell

| Remote code execution | `curl \| sh`, `wget \| bash`, `curl \| python` |

严重 README.md:441exec-pipe-to-shell

| Remote code execution | `curl \| sh`, `wget \| bash`, `curl \| python` |

严重 README.md:444cred-paths

| Credential access | `process.env`, `.env`, `apiKey` references |

严重 README.md:481exec-pipe-to-shell

5. Scans for suspicious processes (`curl | sh`, `wget | bash`, `nc -`, `ncat`)

严重 README.md:481exec-pipe-to-shell

5. Scans for suspicious processes (`curl | sh`, `wget | bash`, `nc -`, `ncat`)

严重 README.md:543meta-injection

| identity_hijacking | Attempts to override the agent's role or instructions | "you are now", "ignore previous instructions", "new system prompt" |

严重 README.md:557exec-pipe-to-shell

| remote_code_execution | CRITICAL | BLOCK | `curl \| sh`, `wget \| bash`, `curl \| python` |

严重 README.md:557exec-pipe-to-shell

| remote_code_execution | CRITICAL | BLOCK | `curl \| sh`, `wget \| bash`, `curl \| python` |

严重 README.md:591exec-pipe-to-shell

Before executing high-risk commands (`rm -rf`, `curl | sh`, `eval`, `chmod 777`, credential access, mass messaging, SQL `DROP`/`DELETE`, `git push --force`, con

严重 README.md:805cred-paths

Watches `~/.openclaw/credentials/` and `~/.openclaw/.env` for file changes using filesystem events.

严重 README.md:860cred-paths

Change every API key, token, and password stored in `.env` or the credentials directory. Assume all credentials that were accessible to the agent have been comp

严重 README.md:954cred-paths

**Credential theft.** An attacker or compromised skill reads API keys from `.env` or credential files and exfiltrates them. SecureClaw addresses this with file

严重 skill/configs/dangerous-commands.json:38cred-paths

"curl.*-X\\s+POST.*\\.env", "wget.*--post-file", "nc\\s+-"]

严重 skill/configs/supply-chain-ioc.json:9cred-paths
```
"process\\.env", "fs\\.readFile.*\\.env",
```
严重 skill/configs/supply-chain-ioc.json:9cred-paths
```
"process\\.env", "fs\\.readFile.*\\.env",
```
严重 skill/configs/supply-chain-ioc.json:25cred-paths
```
"~/.openclaw/.env", "~/.openclaw/credentials/",
```
严重 skill/configs/supply-chain-ioc.json:26cred-paths
```
"~/.clawdbot/.env", "~/.clawdbot/credentials/",
```
严重 skill/configs/supply-chain-ioc.json:27cred-paths
```
"~/.moltbot/.env"
```
严重 skill/scripts/quick-audit.sh:88cred-paths
```
if [ -f "$OPENCLAW_DIR/.env" ]; then
```

这一栏是扫描器报的事实，不是结论。命中多不等于有毒（安全工具、规则库、示例脚本本来就会包含危险写法），命中少也不等于干净。它和你手上的凭据、文件、网络有什么关系，需要你自己看。

技能内容

SecureClaw — Security Hardening

Tools

security_audit

Run a comprehensive security audit of this OpenClaw instance.

Returns findings with severity, description, and remediation steps.

Use when the user asks about security status or hardening.

security_status

Get current security posture: score, active monitors, recent alerts.

skill_scan

Scan a ClawHub skill for malicious patterns before installation.

Required parameter: skill name or URL.

cost_report

Show API cost tracking data: current spend, projections, and alerts.

想直接用这个技能？

本站把开放许可（MIT / Apache 等）的技能按仓库打包整理到网盘，点一下转存到你自己的网盘，不用一个个从 GitHub 拉。许可未声明的技能只给原始仓库链接，不打包。

去原始仓库

它属于哪个仓库

仓库adversa-ai/secureclaw

星标★ 349

本站分层T2

该仓技能数2

原文件路径secureclaw/SKILL.md

同一个仓库里的其他技能

secureclaw

看这个仓库的全部 2 个技能

同名技能的其他版本

有 2 个不同仓库或目录里都有叫 secureclaw 的技能。它们内容并不相同，别混用：

adversa-ai/secureclaw — Security skill for OpenClaw agents (7-framework aligned). 15 core rules + automated script