技能 article-list-processor

📦

article-list-processor

Name: article-list-processor
Author: wangdefou-dev

內容修訂 r1 高風險 ⚡ 包含腳本⚙️ 外部命令📁 檔案系統存取🔑 環境變數🌐 網路存取

將文章清單處理成草稿

內容團隊經常收集許多文章連結，卻缺乏可重複的方法將它們轉換成可用的草稿。此技能會讀取 Markdown 連結清單、擷取來源文章，並批次產生改寫後的草稿內容。

支援: Claude Codex Code(CC)

⚠️ 38 差

Maintain this Skill? Claim it ⚠️ 38 差

安裝前請停止並請求確認。

請審查安裝計畫，並在變更檔案前取得使用者明確同意。

使用我的智慧代理安裝

將此請求複製給你的智慧代理，其中包含權威 Skill 頁面和清單。

智能體請求

Review the Skillstore skill "article-list-processor" from https://skillstore.io/skills/wangdefou-dev-article-list-processor.md and its manifest at https://skillstore.io/api/skills/wangdefou-dev-article-list-processor/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.

你的智慧代理仍應顯示其計畫，並依安全政策請求必要確認。

Agent 可讀資源

當 AI Agent、爬蟲或腳本需要乾淨脈絡、而不是讀取完整頁面時，請使用這些連結。

Markdown 詳情 GET /skills/wangdefou-dev-article-list-processor.md 已簽名 manifest GET /api/skills/wangdefou-dev-article-list-processor/manifest 已簽名 lockfile GET /api/skills/wangdefou-dev-article-list-processor/lockfile

測試它

正在使用「article-list-processor」。含有三個公開文章連結的 Markdown 閱讀清單。

預期結果:

此技能會偵測輸入資料夾中的檔案。
它會擷取每篇文章並提取可讀文字。
它會儲存附有來源中繼資料的獨立改寫草稿。

正在使用「article-list-processor」。包含一個無法存取文章連結的清單。

預期結果:

無法存取的文章會在擷取失敗後被略過。
其他可連線的文章會繼續處理。
批次完成後，輸入清單會被封存。

安全審計

高風險

v9 • 2026/7/7 開啟版本化報告

I found no evidence of obfuscated malware or deliberate credential exfiltration. Confirmed risks include project .env and API key access, unrestricted outbound URL fetching, automatic file watching and reads, and an npm script execution path.

已掃描檔案

741

分析行數

複核項目

已忽略的誤報

已確認的安全問題 (6)

高

index.ts:14

Environment file access

const envPath = path.join(projectRoot, '.env');

The code builds a path to the project .env file outside the skill directory. That file is later loaded and can contain credentials, so this is real sensitive file access.

高

index.ts:25

Environment file access

const ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;

The line reads ANTHROPIC_API_KEY from the loaded environment. This gives the community skill runtime access to an API secret.

高

index.ts:26

Environment file access

const ANTHROPIC_BASE_URL = process.env.ANTHROPIC_BASE_URL;

The base URL setting controls where Anthropic client traffic is sent. If a hostile or mistaken value is loaded from .env, requests and possibly credentials can be redirected.

高

EVALUATION.json:15-18

Prompt Injection Attempt Detected

EVALUATION.json includes self-assessment fields that claim safe_to_publish is true and summarize the skill as low risk. This embedded self-approval can bias automated review and should not be trusted as authority.

The file does not directly command the auditor to skip review, but it asserts its own safety in machine-readable audit fields. The prompt-injection guidance flags self-approval claims as suspicious.

中

index.ts:119-185

Untrusted Article Content Enters LLM Prompt

Fetched article titles and body text are inserted directly into the model prompt. A malicious page can include instructions that alter generated output or weaken safety requirements.

The prompt template directly interpolates articleTitle and articleContent into the user message. There is no visible instruction boundary telling the model to treat fetched content as untrusted data.

中

index.ts:63-66 index.ts:80-84

Unrestricted URL Fetching From Markdown Input

Any parsed link that starts with http is fetched without host validation, scheme restrictions beyond the prefix, or private-network blocking. An untrusted input list can make the runner contact internal or unwanted endpoints.

The parser accepts any link beginning with http, and fetchArticleContent passes that URL directly to fetch. I found no allowlist, DNS/IP filtering, or localhost/private address guard.

能力複核項目 (16)

這些是真實的本機能力，對此技能而言可能是預期行為，因此需要複核，但不會按已確認的惡意行為計分。

高

index.ts:10

Node.js child_process module

import { spawn } from 'child_process';

The child_process import enables the verification routine that spawns npm later in the file. This gives the skill external command execution capability even though the command is hardcoded.

高

index.ts:198

Process spawn

const child = spawn('npm', ['run', 'skill:verify'], {

The skill runs npm run skill:verify after generating content, with shell execution enabled nearby. A package script can execute arbitrary local commands, so this is a real execution risk despite fixed arguments.

高

index.ts:25

Environment variable access (dot notation)

const ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;

The skill reads ANTHROPIC_API_KEY from the environment. That is expected for model access, but marketplace users must trust the skill with a live API credential.

高

index.ts:25

Environment variable object

const ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;

The skill reads ANTHROPIC_API_KEY from the environment. That is expected for model access, but marketplace users must trust the skill with a live API credential.

高

index.ts:23

dotenv library

dotenv.config({ path: envPath, override: true });

dotenv.config loads the project .env file with override enabled. This is sensitive because the skill runtime receives project secrets, even if it mainly uses Anthropic settings.

高

index.ts:25

Generic API/secret keys

const ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;

ANTHROPIC_API_KEY is read and passed into the Anthropic client. This is legitimate configuration, but it is still direct handling of a high-value API secret by community skill code.

高

index.ts:41

Generic API/secret keys

apiKey: ANTHROPIC_API_KEY || 'dummy',

ANTHROPIC_API_KEY is read and passed into the Anthropic client. This is legitimate configuration, but it is still direct handling of a high-value API secret by community skill code.

中

index.ts:80

Fetch API call

const response = await fetch(url, {

fetch(url) requests article links parsed from user-provided Markdown without an allowlist or private-network block. This can reach arbitrary HTTP targets from the user environment.

中

index.ts:6

HTTP client library

import fetch from 'node-fetch';

node-fetch is imported and used to retrieve URLs parsed from local Markdown input. This creates outbound network capability that should be reviewed for SSRF and data-flow boundaries.

中

index.ts:230

Node.js fs operations

const fileContent = fs.readFileSync(filePath, 'utf-8');

The skill synchronously reads filePath values delivered by the watcher. The path is expected to be under local_inputs, but there is no realpath or symlink validation before reading.

中

index.ts:230

Synchronous file operations

const fileContent = fs.readFileSync(filePath, 'utf-8');

The skill synchronously reads filePath values delivered by the watcher. The path is expected to be under local_inputs, but there is no realpath or symlink validation before reading.

中

index.ts:9

File system watching

import chokidar from 'chokidar';

The chokidar dependency is used to monitor an input directory and trigger automatic processing. The watched scope is limited, but it is still a real file observation surface.

中

index.ts:315

File system watching

const watcher = chokidar.watch(INPUT_DIR, {

The watcher monitors local_inputs and processes existing files because ignoreInitial is false nearby. This automatic file processing is documented, but it can surprise users if stale files are present.

中

index.ts:26

Environment variable access (dot notation)

const ANTHROPIC_BASE_URL = process.env.ANTHROPIC_BASE_URL;

The skill reads ANTHROPIC_BASE_URL and passes it to the SDK. A custom endpoint can redirect model traffic if the environment is misconfigured or tampered with.

中

index.ts:26

Environment variable object

const ANTHROPIC_BASE_URL = process.env.ANTHROPIC_BASE_URL;

The skill reads ANTHROPIC_BASE_URL and passes it to the SDK. A custom endpoint can redirect model traffic if the environment is misconfigured or tampered with.

中

index.ts:3

dotenv library

import dotenv from 'dotenv';

dotenv is imported and used to load project environment variables. Loading .env gives the skill access to local configuration and potential secrets.

風險因素

⚡ 包含腳本 (5)

EVALUATION.json:18 EVALUATION.json:21 EVALUATION.json:24 EVALUATION.json:42 index.ts:63

⚙️ 外部命令 (48)

📁 檔案系統存取 (25)

🔑 環境變數 (18)

EVALUATION.json:72 EVALUATION.json:75 EVALUATION.json:78 EVALUATION.json:171 EVALUATION.json:174 EVALUATION.json:72 EVALUATION.json:162 EVALUATION.json:339 index.ts:25 index.ts:26 index.ts:27 index.ts:25 index.ts:26 index.ts:27 index.ts:3 index.ts:23 index.ts:25 index.ts:41

🌐 網路存取 (6)

index.ts:80 index.ts:6 index.ts:82 SKILL.md:30 SKILL.md:31 SKILL.md:32

審計單位: codex 檢視審計紀錄 →

分享與引用此報告

分享版本化評估報告、中性徽章、嵌入卡片與引用。Skillstore 只報告證據，不判斷該 Skill 是否安全。

開啟版本化報告

複製報告連結

https://skillstore.io/skills/wangdefou-dev-article-list-processor/audits/9?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown 徽章

[![Skillstore security assessment](https://skillstore.io/badges/skills/wangdefou-dev-article-list-processor/security.svg)](https://skillstore.io/skills/wangdefou-dev-article-list-processor?utm_source=security_passport_badge)

HTML 徽章

<a href="https://skillstore.io/skills/wangdefou-dev-article-list-processor?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/wangdefou-dev-article-list-processor/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

嵌入卡片

<iframe src="https://skillstore.io/embed/skills/wangdefou-dev-article-list-processor.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

學術引用 (APA · BibTeX · CFF)

APA 引用

wangdefou-dev. (2026). article-list-processor security audit report (audit version 9) [Author version unspecified]. Skillstore. https://skillstore.io/skills/wangdefou-dev-article-list-processor/audits/9

BibTeX 引用

@techreport{wangdefou-dev-wangdefou-dev-article-list-processor-2026,
  author = {wangdefou-dev},
  title = {article-list-processor security audit report (audit version 9)},
  institution = {Skillstore},
  year = {2026},
  number = {9},
  url = {https://skillstore.io/skills/wangdefou-dev-article-list-processor/audits/9},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "article-list-processor security audit report (audit version 9)"
version: "unspecified"
type: report
authors:
  - name: "wangdefou-dev"
date-released: "2026-07-07"
url: "https://skillstore.io/skills/wangdefou-dev-article-list-processor/audits/9"
identifiers:
  - type: other
    value: "skillstore:wangdefou-dev-article-list-processor:audit:9"
    description: "Skillstore immutable audit report identifier"

Skillstore 評分

評分依據證據置信度: 中

架構

可維護性

內容

社群

規範符合度

你可以打造什麼

將閱讀清單轉換成草稿

將儲存的文章連結轉換成供後續編輯與發布使用的初稿。

批次處理研究連結

將精選的網路文章清單處理成內容行事曆可用的結構化草稿素材。

準備社群評論

從多篇來源文章產生偏向觀點評論的改寫內容，供編輯審閱。

試試這些提示

處理我的連結清單

處理我放在本機輸入資料夾中的 Markdown 文章清單，並為每個有效連結建立草稿貼文。

測試工作流程

以 mock mode 執行文章清單工作流程，並確認我的 Markdown 連結已正確偵測。

審閱產生的草稿

審閱產生的文章草稿，檢查事實準確性、語氣、來源標註與重用風險。

稽核批次設定

在我執行大型批次前，檢查文章清單工作流程設定、URL 輸入、輸出路徑與驗證步驟。

最佳實務

僅使用你有權擷取與轉換的 URLs。
先執行小批次，以確認輸出品質與來源處理方式。
審閱產生的草稿，檢查標註、準確性與著作權風險。

避免

未經許可，不要處理私人、付費牆或需登入保護的文章。
不要將敏感的內部 URLs 放入輸入 Markdown 檔案。
未經人工編輯審閱，不要發布產生的草稿。

常見問題

此技能使用什麼輸入格式？

它預期使用 Markdown 檔案，且連結需寫成帶標題的 Markdown 連結。

它需要網際網路存取嗎？

是。它會擷取來源文章頁面，並將 prompts 傳送至 Anthropic-compatible API endpoint。

產生的草稿會儲存在哪裡？

此實作會將產生的 Markdown 檔案寫入已設定的專案輸出目錄下。

它可以一次處理很多連結嗎？

可以。它會從一個輸入檔案處理多個連結，並行限制為兩個。

它會驗證產生的內容嗎？

會。在成功生成後，它會從專案根目錄執行硬編碼的 npm verification script。

它對敏感來源安全嗎？

不適合。請避免使用私人 URLs，因為擷取的內容可能會被傳送至遠端模型 endpoint。

開發者詳細資訊

作者

wangdefou-dev

授權

MIT

Skillstore 修訂

版本提示

作者未聲明版本。

儲存庫

https://github.com/wangdefou-dev/defou-workflow-agent/tree/main/skills/article-list-processor

引用

36e07d5e13068e5be64447e8f20b427cf2cbd21a

維護新鮮度

2026/7/18

使用量

16 次下載 · 206 次瀏覽

檔案結構

📄 EVALUATION.json

📄 index.ts

📄 SKILL.md