automatic-skill 安全扫描报告 — 可信 | ClawSafe

5 /100

automatic-skill

Automatic Skill — 每日自动调研、设计、制作、测试并发布新 Skill 的元技能流水线

This is a legitimate meta-skill pipeline orchestrator with no malicious behavior. All scripts are pure prompt-generators that output text instructions for agents; they perform no dangerous operations themselves. Credential access (GITHUB_TOKEN, CLAWHUB_TOKEN) is fully declared in SKILL.md metadata and used solely for their stated purpose (GitHub commit/push and clawHub API publish). No hidden functionality, credential harvesting, or data exfiltration was found.

技能名称automatic-skill

分析耗时65.3s

引擎pi

✓

可以安装

This skill is safe to use. The declared environment variables (GITHUB_TOKEN, CLAWHUB_TOKEN) are necessary and appropriate for its publish-to-GitHub-and-clawHub purpose. No action required.

安全发现 2 项

严重性	安全发现	位置
低危	SKILL.md lists 9 stages but pipeline.js implements 10 文档欺骗 The SKILL.md documentation describes a '9-stage pipeline' (Research→Design→Create→Review→Self-Run→Self-Check→Upload→Verify→Final Review) as 9 rows in the table. However, pipeline.js includes an additional 'SEO' stage (Stage 3) between Design and Create, making it effectively 10 stages. This is a minor documentation inconsistency. \| 3 \| 制作 Create \| `create.js <design-file>` \| 按设计生成 SKILL.md、scripts、package.json 等全部文件 \| → Update SKILL.md to correctly reflect the 10-stage pipeline including SEO as Stage 3.	`SKILL.md:38-46 vs scripts/pipeline.js:52:38`
低危	SKILL.md documentation table has wrong stage numbers 文档欺骗 SKILL.md maps Stage 3 to Create and Stage 4 to Review, but pipeline.js maps Stage 3 to SEO and Stage 4 to Create, Stage 5 to Review, etc. This means every stage number from 3 onwards is off by one in the documentation. `\| 3 \| 制作 Create \| create.js \| ... \|` → Align the SKILL.md stage table with the actual pipeline.js stage ordering.	`SKILL.md:38-46 vs scripts/pipeline.js:52:38`

资源类型	声明权限	推断权限	状态	证据
文件系统	`READ`	`READ`	✓ 一致	Scripts read from data/ JSON files and write pipeline state — all documented in …
网络访问	`READ`	`READ`	✓ 一致	upload.js:90 — curl to api.clawhub.io; verify-upload.js:73 — GET api.clawhub.io.…
命令执行	`NONE`	`NONE`	—	No direct shell:WRITE found. Scripts output prompts instructing agents to run gi…
环境变量	`READ`	`READ`	✓ 一致	GITHUB_TOKEN, GITHUB_REPO, CLAWHUB_TOKEN read and used solely for their document…
技能调用	`NONE`	`NONE`	—	No skill-to-skill invocation.
剪贴板	`NONE`	`NONE`	—	No clipboard access.
浏览器	`NONE`	`NONE`	—	No browser access.
数据库	`NONE`	`NONE`	—	No database access.

2 项发现

🔗

中危外部 URL 外部 URL

https://api.clawhub.io/v1/skills/publish

scripts/upload.js:90

🔗

中危外部 URL 外部 URL

https://api.clawhub.io/v1/skills/$

scripts/verify-upload.js:73

目录结构

19 文件 · 89.5 KB · 2381 行

JavaScript 14f · 2121L Markdown 1f · 174L JSON 4f · 86L

├─ ▾ 📁 data

│ ├─ 📋 current-pipeline.json JSON 21L · 901 B

│ └─ 📋 pipeline-log.json JSON 29L · 856 B

├─ ▾ 📁 scripts

│ ├─ 📜 create.js JavaScript 139L · 6.3 KB

│ ├─ 📜 daily-pipeline.js JavaScript 85L · 3.5 KB

│ ├─ 📜 design.js JavaScript 175L · 5.5 KB

│ ├─ 📜 final-review.js JavaScript 208L · 7.1 KB

│ ├─ 📜 pipeline.js JavaScript 159L · 6.7 KB

│ ├─ 📜 push-toggle.js JavaScript 71L · 2.1 KB

│ ├─ 📜 research.js JavaScript 113L · 4.4 KB

│ ├─ 📜 review.js JavaScript 153L · 7.0 KB

│ ├─ 📜 self-check.js JavaScript 155L · 5.4 KB

│ ├─ 📜 self-run.js JavaScript 146L · 5.2 KB

│ ├─ 📜 seo.js JavaScript 193L · 6.6 KB

│ ├─ 📜 status.js JavaScript 138L · 5.4 KB

│ ├─ 📜 upload.js JavaScript 203L · 7.4 KB

│ └─ 📜 verify-upload.js JavaScript 183L · 6.6 KB

├─ 📋 _meta.json JSON 6L · 126 B

├─ 📋 package.json JSON 30L · 1.2 KB

└─ 📝 SKILL.md Markdown 174L · 7.3 KB

依赖分析 1 项

包名	版本	来源	已知漏洞	备注
`(none)`	`N/A`	npm	否	package.json has no external dependencies — all scripts use only Node.js built-in modules (fs, path)

安全亮点

✓ No base64-encoded execution, eval(), or Function() with user-controlled strings

✓ No credential harvesting or exfiltration — tokens used only for declared publish purpose

✓ No reverse shell, C2 communication, or hidden network exfiltration

✓ No access to sensitive paths (~/.ssh, ~/.aws, .env, /etc/passwd)

✓ No obfuscation techniques (base64 piping, atob, etc.)

✓ No supply chain risks — package.json has no external dependencies

✓ Security checklist is included in review.js and enforces: no hardcoded API keys, input validation, no eval()

✓ Dry-run mode skips all network operations — safe for testing

✓ Lock file prevents duplicate pipeline runs

✓ Pipeline state is stored locally in data/ JSON files, not transmitted externally