據 AlphaMoat 統計,截至 2026-10,Skill Scanner 累計下載 93,244 次、獲 4 個星標,按累計下載量在本次參與排名的 10,914 個 AI 技能中排名第 268 位,在「運維安全」分類(502 個)中排名第 5 位。
列表預設按累計下載量排序,也可按星標排序;詳情頁排名按累計下載量計算。
| 平台 | 累計下載 | 星標 | 評論 |
|---|---|---|---|
| ClawHub ↗ | 1049 | - | - |
| SkillHub ↗ | 9.3萬 | 4 | — |
各平台的計數口徑不同,統計時間不一致。
用於安裝技能、包或外掛的安全檢查。請在執行任何 npm install、openclaw plugins install、clawhub install 或類似安裝命令之前使用。
| 月份 | 累計下載 | 環比 | 星標 |
|---|---|---|---|
| 2026-10 | 9.3萬 | — | 4 |
| 2026-09 | 2萬 | — | 2 |
以下是作者釋出的原始檔案,按純文字顯示,本站未做修改或稽核;其中的連結不可點選,命令請核實後再執行。抓取於 2026-11-20。
--- name: skill-security description: Security checks for installing skills, packages, or plugins. Use BEFORE any `npm install`, `openclaw plugins install`, `clawhub install`, or similar install commands. Also use when reviewing a newly installed skill before first use. Triggered by any install, add, or package addition request. --- # Skill Security Run these checks before installing ANY skill, package, or plugin. Always warn the user before proceeding. ## Pre-Install Checklist
### 1. Source Verification - Is it from clawhub.com (vetted) or a random npm package? - Who's the author? Do they have other packages / reputation? - Is there a GitHub repo? Check for recent commits, open issues, maintainer activity. ### 2. Popularity Check - `npm info <package>` — check weekly downloads, last publish date, version history - Low downloads (< 100/week) + recent publish = higher risk ### 3. Dependency Audit - `npm info <package> dependencies` — how many deps does it pull in? - More dependencies = larger attack surface - Flag packages with 50+ transitive dependencies ### 4. Lifecycle Scripts (HIGH RISK) - Check for `preinstall`, `install`, `postinstall` scripts — these run arbitrary code - `npm info <package> scripts` or inspect `package.json` - pnpm blocks these by default; npm does NOT - If lifecycle scripts exist, flag it explicitly to the user ### 5. Scan After Install - `npm audit` after install to catch known vulnerabilities ### 6. Check for Dynamic Content - Search the skill code for URLs that are fetched at runtime - Skills that download and execute content from external endpoints can change behavior after install ## Red Flags — Stop and Ask - 🚩 Package published very recently with no history - 🚩 Maintainer has no other packages or reputation - 🚩 Package name similar to a popular one (typosquatting: e.g., `reqeust` vs `request`) - 🚩 Requests permissions beyond what it claims to do - 🚩 No GitHub repo, or repo is empty/suspicious - 🚩 Postinstall scripts that download from unknown URLs - 🚩 **Dynamic content fetching** — skill calls external URLs at runtime (2.9% of ClawHub skills do this; payload can change after install) - 🚩 **Base64 in install instructions** — ClickFix social engineering pattern (fake errors → paste base64 command → malware) - 🚩 **New uploader, bulk uploads** — single user uploading many skills rapidly (ClawHavoc: 354 skills from one account) - 🚩 **Skill references credential paths** — `~/.openclaw/credentials/`, `~/.clawdbot/.env`, `.env` files ## Core File Protection Skills are NEVER allowed to modify these files without explicit user approval: - `SOUL.md` — agent identity - `AGENTS.md` — agent rules - `IDENTITY.md` — agent metadata - `USER.md` — user's personal info - `MEMORY.md` or `memory/*.md` — agent memories - `TOOLS.md` — infrastructure notes ## Data Exfiltration Checks After installing a skill, before running it: 1. **Read SKILL.md** — understand what the skill does 2. **Check file paths** — does it reference paths outside its own directory and workspace? 3. **Check exec commands** — does it `curl` to unknown domains? 4. **Check write/edit calls** — are target paths outside `workspace/`? 5. **Check for credential access** — `.env`, `~/.ssh/`, `~/.gnupg/`, API keys 6. **Check for dynamic content** — does it fetch and execute content from external URLs at runtime? (Snyk: 2.9% of ClawHub skills do this) ## ClawHub-Specific Checks When installing from ClawHub: - **Check uploader history** — is this a new account? Do they have other skills? Bulk uploads from unknown accounts = red flag (ClawHavoc: 354 skills from one malicious account) - **Check skill stars/reviews** — community feedback is a signal, not proof - **Check VirusTotal** — OpenClaw has a VirusTotal partnership; check if the skill has been scanned - **Verify semantic versioning** — legitimate skills typically have version history and changelogs - **Search for security reports** — search web for "[skill-name] malicious" before installing ## Known Attack Campaigns (as of March 2026) Reference for identifying patterns: - **ClawHavoc** — 300+ coordinated skills, ClickFix social engineering, downloads Atomic Stealer (AMOS) - **AuthTool** — dormant payload, activates on specific natural language prompts, establishes reverse shell - **Hidden Backdoor** — fake "Apple Software Update" during install, encrypted tunnel to attacker - **Credential Exfiltration** — targets `~/.clawdbot/.env` and `~/.openclaw/credentials/` for API keys ## Red Flags — Skill Behavior - 🚩 Skill reads `.env` or credential files - 🚩 Skill makes network requests to unfamiliar domains - 🚩 Skill writes to paths outside workspace - 🚩 Skill modifies core files (SOUL, AGENTS, MEMORY, USER, IDENTITY, TOOLS) - 🚩 Skill sends data to external URLs not part of stated purpose ## What To Tell The User Before installing, give a brief summary: > "⚠️ Installing [package]: [downloads/week], [last updated], [dep count] deps, [lifecycle scripts?]. Looks [clean/sketchy] — proceed?" If red flags found: > "🚩 Flags on [package]: [list issues]. Want me to proceed anyway?"
鍾馗.Skill——會更新漏洞庫的安全審查專家。直來直去、快刀斬亂麻,三層審查(靜態審計/行為模擬/供應鏈溯源)覆蓋12類風險,輸出結構化安全裁定(✅乾淨/⚠️可疑/🚫惡意)。Use when 使用者說"審查這個Skill"、"安全檢查"、"鍾馗看下"、"審一下"、"查一下這個skill"、安裝Skill前的安全評估、或需要審計SKILL.md的惡意載荷。
在使用技能包之前做安全掃描,並按結論決定放行或攔截的技能。
為 OpenClaw 配置 OpenRouter 免費模型與回退鏈的工具。
Parse, search, and analyze application logs across formats. Use when debugging from log files, setting up structured logging, analyzing error patterns, correlating events across services, parsing stack traces, or monitoring log output in real time.
網路安全與SRC漏洞掘金實戰技能。融合「系統學習路徑」與「SRC實戰掘金」雙模式: ①學習模式=法律→基礎→工具→靶場→攻防→就業全流程; ②實戰模式=直接指導在騰訊TSRC、阿里ASRC、百度BSRC等平台挖掘高價值漏洞變現。 當用戶提出以下問題時觸發: "我想學網路安全怎麼入門""SRC漏洞怎麼挖能賺錢""滲透測試需要學什麼" "Web安全漏洞原理是什麼""怎麼準備CTF比賽""護網面試會問什麼" "雲安全怎麼學""挖漏洞賺錢的方法""埠掃描指令碼怎麼寫" "SQL隱碼攻擊怎麼檢測""子域名怎麼列舉""XSS漏洞怎麼找" "靶場怎麼搭建""Kali Linux怎麼用""Metasploit怎麼入門" "等保測評是什麼""程式碼審計怎麼做""內網滲透怎麼學" "資訊洩露漏洞怎麼挖""越權漏洞怎麼測試""邏輯漏洞挖掘技巧" "簡歷怎麼寫才能進安全公司""安全行業面試題有哪些" "Burp Suite怎麼用""DVWA怎麼安裝""Pikachu靶場搭建" "Vulhub漏洞復現""網路安全法有什麼規定""CTF奪旗賽怎麼參加" "AWD比賽是什麼""應急響應怎麼做""攻擊溯源怎麼搞" "資料安全怎麼保障""公有云安全怎麼防護""私有云攻防怎麼做" "雲原生安全是什麼"。
這個技能特別簡單——你只需要呼叫它,給一個網站域名,它就會自動幫你跑一遍滲透測試,找出這個網站存在哪些安全漏洞。 不需要你配環境、不用你手動發包,一切自動化。測完直接出結果,告訴你哪裡有問題、怎麼修。 它的願景是:讓每個人都能成為白帽駭客——只要輸入一個域名,就能知道自己的網站安不安全。一切只為授權測試,只為發現漏洞、修復風險,讓網際網路更安全。
Security-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope, and suspicious patterns.
# 「礪刃」 授權場景下,將紅隊實戰能力轉化為可直接提交的SRC有效戰果。 面向**授權滲透測試、補天、廠商SRC及眾測平台**打造的一體化挖洞工具鏈。 融合 pentest‑lyan 工程化滲透方法論與 src‑vuln‑hunting 實戰提交規範;支援模型自主威脅建模,本地持久化任務狀態,全流程支援斷點續測。 ## 核心能力 - 12維繫統化威脅建模:覆蓋資料流、許可權邊界、資源歸屬、狀態變更、客戶端可控輸入、併發競爭場景、輸出渲染、認證與會話機制、SSRF風險點、各類注入面、檔案操作、業務邏輯缺陷,完整覆蓋業務Web常見攻擊面。 - JS多源深度解析:獨立JS檔案、頁面內聯指令碼、外部資源引用、頁面跳轉連結、路徑推理、響應動態注入六大采集渠道,自動提取並梳理全部後端介面。 - 三層確認校驗門閂機制:依次校驗訪問可達性、漏洞存在性、可利用性,三項全部透過方可標記為 confirmed,規避“返回200即判定漏洞”的常見誤報問題。 - SRC標準化報告輸出(補天風格):自動整理原始請求/響應資料包、危害場景證明、不少於2條可落地修復建議,產出內容可直接用於漏洞報送。 - 白盒審計聯動:對接 src‑6k 白盒審計能力,實現介面探測到原始碼審計的打通,完整還原漏洞利用鏈路。 ## 適用場景 1. 獲得正式授權的目標:常規滲透測試、輕量Web應用安全評估 2. 補天、各廠商SRC、商業眾測任務:減少重複開發,開箱即用,對齊平台提交標準 3. 團隊協同作業:任務狀態按模組持久化儲存,支援斷點續測,避免工作進度丟失 ## 合規邊界 僅可在**已獲取書面授權**的範圍內開展測試;恪守最小驗證原則,做好測試樣本管控。 預授權介面、驗證碼、隨機金鑰類物件不隨意標記高危;僅驗證訪問可達性,不隨意拔高風險等級。 --- ### 【SkillHub卡片簡介|單行精簡版,直接複製用】 「礪刃」——授權環境下把紅隊手法轉化為可提交SRC戰果。面向授權滲透、補天及廠商SRC眾測的一體化挖洞工具鏈。內建12維威脅建模、JS多源解析、三層校驗防誤報,輸出補天規範漏洞報告,支援斷點續測與白盒聯動;嚴格恪守測試授權邊界,適配個人與團隊挖洞作業。
資料月份:2026-10 · 下載/星標數聚合自公開技能註冊庫(ClawHub、SkillHub),按月更新。口徑說明見資料方法。