中文 EN

扫描报告

扫描新文件

⚠ 低风险 — 风险评分 22/100
上次扫描：21 小时前 重新扫描
22 /100
contentclaw
Turn papers, podcasts, and case studies into publish-ready social posts, infographics, and diagrams
A legitimate content generation skill with well-scoped permissions, no credential theft, and no hidden malicious behavior; the only notable concern is a documented curl|sh installation pattern for a known tool (uv).
技能名称contentclaw
分析耗时54.3s
引擎pi
✓
可以安装
Approve for use. Consider pinning dependency versions in pyproject.toml to reduce supply chain risk. The curl|sh command is a documented installation step and involves a well-known tool (Astral's uv), but users should be aware of the pattern.

安全发现 4 项

严重性 安全发现 位置
中危
Documented curl|sh installation pattern 代码执行
SKILL.md instructs `curl -LsSf https://astral.sh/uv/install.sh | sh` to install the uv package manager. While documented and involving a known reputable tool (Astral), pipe-to-bash remains a high-risk pattern that should be avoided when possible.
curl -LsSf https://astral.sh/uv/install.sh | sh
→ Prefer Homebrew (macOS) or pipx installation methods which are safer. If curl|sh must be used, always review the script at the URL first.
 SKILL.md:56
低危
Unpinned dependency versions 供应链
All dependencies in pyproject.toml use >= lower bounds without upper limits, allowing any version including potentially malicious updates to be pulled.
httpx>=0.27, pymupdf>=1.24, readabilipy>=0.2, playwright>=1.49, fal-client>=0.5, exa-py>=1.0
→ Pin to known-good versions or use a lock file (uv.lock) to ensure reproducible and secure builds.
 pyproject.toml:10
低危
False-positive IP address flagged in pre-scan 敏感访问
The pre-scan flagged '131.0.0.0' in browser.py:17 as a hardcoded IP. Inspection reveals it is a User-Agent browser version string (Chrome/131.0.0.0), not an IP address. This is a false positive.
USER_AGENT = "Mozilla/5.0 ... Chrome/131.0.0.0 Safari/537.36"
→ No action needed; pre-scan rule should be refined to distinguish User-Agent version strings from IP addresses.
 scripts/browser.py:17
提示
SKILL.md declares allowed-tools but capability mapping is mostly aligned 文档欺骗
allowed-tools declares Bash, Read, Write, Edit, Glob, Grep, AskUserQuestion. The skill actually uses shell:WRITE (uv run), filesystem:READ/WRITE (scoped to BASE_DIR), network:READ (API calls), environment:READ (scoped .env), browser:READ (Playwright). No Edit tool usage was found in the scripts.
allowed-tools: Bash, Read, Write, Edit, Glob, Grep, AskUserQuestion
→ Update allowed-tools to accurately reflect browser usage (Playwright) or document it as a prerequisite.
 SKILL.md:34
资源类型声明权限推断权限状态证据
文件系统 READ WRITE ✓ 一致 SKILL.md declares Read/Write/Edit tools but Write usage is scoped to BASE_DIR co…
网络访问 READ READ ✓ 一致 Makes outbound API calls to fal.ai and exa.ai (documented), no inbound listeners
命令执行 WRITE WRITE ✓ 一致 Uses uv run for subprocess execution; curl|sh documented in SKILL.md line 56
环境变量 NONE READ ✓ 一致 Scoped .env loading of FAL_KEY and EXA_API_KEY in env.py, discover_topics.py, ge…
技能调用 NONE READ ✓ 一致 Skill invokes itself via recipe execution pattern; no privilege escalation
剪贴板 NONE NONE — No clipboard access found
浏览器 NONE READ ✓ 一致 Playwright headless browser used for extraction; declared in install prerequisit…
数据库 NONE NONE — No database access
1 严重 1 高危 20 项发现
💀
严重 危险命令 危险 Shell 命令
curl -LsSf https://astral.sh/uv/install.sh | sh
SKILL.md:56
📡
高危 IP 地址 硬编码 IP 地址
131.0.0.0
scripts/browser.py:17
🔗
中危 外部 URL 外部 URL
https://docs.astral.sh/uv/
SKILL.md:53
🔗
中危 外部 URL 外部 URL
https://astral.sh/uv/install.sh
SKILL.md:56
🔗
中危 外部 URL 外部 URL
https://www.reddit.com/r/HowToAIAgent/comments/1rfo5i7/i_opensourced_my_kindle_publishing_pipeline_with/
recipes/demo-diagram-breakdown.yaml:50
🔗
中危 外部 URL 外部 URL
https://www.linkedin.com/posts/romejgeorgio_the-start-of-machine-to-machine-marketing-activity-7430322995220459520-gU7Q
recipes/paper-breakdown-insight.yaml:49
🔗
中危 外部 URL 外部 URL
https://x.com/omarsar0/status/2030403147588604376
recipes/paper-breakdown-insight.yaml:63
🔗
中危 外部 URL 外部 URL
https://x.com/omarsar0/status/2029926242640912429
recipes/paper-breakdown-insight.yaml:64
🔗
中危 外部 URL 外部 URL
https://www.linkedin.com/posts/romejgeorgio_llms-are-curating-reality-x-open-sourcing-activity-7419815247869333504-l-KW
recipes/podcast-insight.yaml:51
🔗
中危 外部 URL 外部 URL
https://www.reddit.com/r/gtmengineering/comments/1q805wm/how_were_personalising_cold_emails_at_scale_in/
recipes/reddit-short-case-study.yaml:49
🔗
中危 外部 URL 外部 URL
https://www.reddit.com/r/aiagents/comments/1rctr3d/agents_are_getting_more_powerful_every_day_here/
recipes/what-you-might-have-missed.yaml:52
🔗
中危 外部 URL 外部 URL
https://www.reddit.com/r/aiagents/comments/1r6jo63/its_been_a_big_week_for_agentic_ai_here_are_10/
recipes/what-you-might-have-missed.yaml:53
🔗
中危 外部 URL 外部 URL
https://arxiv.org/abs/2401.04088
tests/test_extract.py:33
🔗
中危 外部 URL 外部 URL
https://arxiv.org/pdf/2401.04088
tests/test_extract.py:34
🔗
中危 外部 URL 外部 URL
https://reddit.com/r/test/comments/abc
tests/test_extract.py:35
🔗
中危 外部 URL 外部 URL
https://x.com/user/status/123
tests/test_extract.py:37
🔗
中危 外部 URL 外部 URL
https://twitter.com/user/status/123
tests/test_extract.py:38
🔗
中危 外部 URL 外部 URL
https://youtube.com/watch?v=abc
tests/test_extract.py:41
🔗
中危 外部 URL 外部 URL
https://lilianweng.github.io/posts/2024-02-05-human-data-quality/
tests/test_extract.py:55
🔗
中危 外部 URL 外部 URL
https://thisdoesnotexist.example.com
tests/test_extract.py:74

目录结构

46 文件 · 101.9 KB · 2923 行
Markdown 15f · 1241L Python 8f · 1172L YAML 22f · 487L TOML 1f · 23L
├─ ▾ 📁 agents
│ ├─ 📝 breakdown.md Markdown 51L · 1.8 KB
│ ├─ 📝 caption.md Markdown 39L · 1.4 KB
│ ├─ 📝 case-study.md Markdown 56L · 2.0 KB
│ ├─ 📝 diagram.md Markdown 59L · 2.0 KB
│ ├─ 📝 infographic.md Markdown 58L · 2.0 KB
│ ├─ 📝 insight-post.md Markdown 53L · 2.2 KB
│ ├─ 📝 poster.md Markdown 51L · 1.7 KB
│ ├─ 📝 reddit-human.md Markdown 72L · 3.8 KB
│ └─ 📝 roundup.md Markdown 49L · 1.8 KB
├─ ▾ 📁 brand-graphs
│ └─ ▾ 📁 templates
│ ├─ ▾ 📁 ai-ml
│ │ ├─ 📋 audience.yaml YAML 13L · 313 B
│ │ ├─ 📋 feedback.yaml YAML 1L · 13 B
│ │ ├─ 📋 identity.yaml YAML 6L · 218 B
│ │ ├─ 📋 strategy.yaml YAML 9L · 220 B
│ │ └─ 📋 visual.yaml YAML 2L · 49 B
│ ├─ ▾ 📁 dev-tools
│ │ ├─ 📋 audience.yaml YAML 13L · 293 B
│ │ ├─ 📋 feedback.yaml YAML 1L · 13 B
│ │ ├─ 📋 identity.yaml YAML 6L · 196 B
│ │ ├─ 📋 strategy.yaml YAML 8L · 214 B
│ │ └─ 📋 visual.yaml YAML 2L · 49 B
│ └─ ▾ 📁 saas-b2b
│ ├─ 📋 audience.yaml YAML 13L · 311 B
│ ├─ 📋 feedback.yaml YAML 1L · 13 B
│ ├─ 📋 identity.yaml YAML 6L · 228 B
│ ├─ 📋 strategy.yaml YAML 8L · 194 B
│ └─ 📋 visual.yaml YAML 2L · 49 B
├─ ▾ 📁 recipes
│ ├─ 📋 _schema.yaml YAML 44L · 2.0 KB
│ ├─ 📋 demo-diagram-breakdown.yaml YAML 64L · 1.7 KB
│ ├─ 📋 news-event-poster.yaml YAML 59L · 1.4 KB
│ ├─ 📋 paper-breakdown-insight.yaml YAML 64L · 1.9 KB
│ ├─ 📋 podcast-insight.yaml YAML 63L · 1.7 KB
│ ├─ 📋 reddit-short-case-study.yaml YAML 49L · 1.4 KB
│ └─ 📋 what-you-might-have-missed.yaml YAML 53L · 1.6 KB
├─ ▾ 📁 references
│ ├─ 📝 brand.md Markdown 36L · 1.3 KB
│ ├─ 📝 create-recipe.md Markdown 29L · 1.5 KB
│ ├─ 📝 run-recipe.md Markdown 59L · 2.7 KB
│ └─ 📝 topics.md Markdown 20L · 700 B
├─ ▾ 📁 scripts
│ ├─ ▾ 📁 extractors
│ │ └─ 🐍 extract.py Python 295L · 8.4 KB
│ ├─ 🐍 browser.py Python 40L · 1.2 KB
│ ├─ 🐍 discover_topics.py Python 223L · 6.6 KB
│ ├─ 🐍 env.py Python 22L · 733 B
│ └─ 🐍 generate_image.py Python 287L · 9.4 KB
├─ ▾ 📁 tests
│ ├─ 🐍 test_extract.py Python 85L · 3.1 KB
│ ├─ 🐍 test_image_gen.py Python 91L · 2.7 KB
│ └─ 🐍 test_recipes.py Python 129L · 4.4 KB
├─ 📄 pyproject.toml TOML 23L · 431 B
├─ 📝 SKILL.md Markdown 508L · 21.0 KB
└─ 📝 TODOS.md Markdown 101L · 5.4 KB

依赖分析 6 项

包名版本来源已知漏洞备注
httpx >=0.27 pip 否 Version not pinned
pymupdf >=1.24 pip 否 Version not pinned
readabilipy >=0.2 pip 否 Version not pinned; less widely-used package
playwright >=1.49 pip 否 Version not pinned
fal-client >=0.5 pip 否 Version not pinned
exa-py >=1.0 pip 否 Version not pinned

安全亮点

✓ Scoped .env loading — only FAL_KEY and EXA_API_KEY are loaded, preventing credential over-exposure
✓ Strong file scoping — SKILL.md explicitly restricts reads/writes to BASE_DIR only
✓ No credential theft — no iteration over os.environ, no access to ~/.ssh, ~/.aws, or other sensitive paths
✓ No exfiltration — extracted content is processed locally; external API calls only send condensed specs and search queries
✓ No obfuscation — no base64-encoded strings, eval(), or dynamic code generation
✓ No persistence mechanisms — no cron jobs, startup hooks, or backdoor installation
✓ No C2 communication — no hardcoded IPs for command-and-control
✓ Legitimate toolchain — uses well-known, reputable services (fal.ai, exa.ai, Playwright, Astral uv)
✓ Transparent data flow — SKILL.md documents all external API calls and browser automation clearly