Felo Web Fetch Skill

When to Use

Trigger this skill when users want to extract or convert webpage content from a URL:

• Fetch or scrape content from a webpage URL
• Convert webpage content to html, markdown, or text
• Extract specific blocks using CSS selector
• Get article/main text from a link with readability mode
• Tune extraction behavior with crawl mode (fast/fine)
• Pass request details such as cookies, user-agent, timeout

Trigger keywords (examples):

• fetch webpage, scrape URL, fetch page content, web fetch, url to markdown
• Explicit: /felo-web-fetch, "use felo web fetch", "extract this URL with felo"
• Same intent in other languages (e.g. 网页抓取, 提取网页内容) also triggers this skill

Do NOT use this skill for:

• Real-time Q&A search summaries (use felo-search)
• Slide generation tasks (use felo-slides)
• Local file parsing in current workspace

Setup

1. Get API key

1. Visit felo.ai
2. Open Settings -> API Keys
3. Create and copy your API key

2. Configure environment variable

Linux/macOS:

export FELO_API_KEY="your-api-key-here"

Windows PowerShell:

$env:FELO_API_KEY="your-api-key-here"

How to Execute

Option A: Use the bundled script or packaged CLI

Script (from repo):

node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com/article" [options]

Packaged CLI (after npm install -g felo-ai): same options, with short forms allowed:

felo web-fetch -u "https://example.com" [options]
# Short forms: -u (url), -f (format), -t (timeout, seconds), -j (json)

Required parameter:

• --url

Core optional parameters:

• --output-format html|markdown|text
• --crawl-mode fast|fine
• --target-selector "article.main-content"
• --wait-for-selector ".content-ready"

Other key optional parameters:

• --cookie "session_id=xxx" (repeatable)
• --set-cookies-json '[{"name":"sid","value":"xxx","domain":"example.com"}]'
• --user-agent "Mozilla/5.0 ..."
• --timeout 60 (HTTP request timeout in seconds)
• --request-timeout-ms 15000 (API payload timeout in ms)
• --with-readability true
• --with-links-summary true
• --with-images-summary true
• --with-images-readability true
• --with-images true
• --with-links true
• --ignore-empty-text-image true
• --with-cache false
• --with-stypes true
• --json (print full JSON response)

How to write instructions (target_selector + output_format)

When the user wants a specific part of the page or a specific output format, phrase the command like this:

• Output format: "Fetch as text" / "Get markdown" / "Return html" → use --output-format text, --output-format markdown, or --output-format html.
• Target one element: "Only the main article" / "Just the content inside #main" / "Fetch only article.main-content" → use --target-selector "article.main" or the selector they give.

Examples:

# Basic: fetch as Markdown
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com" --output-format markdown

# Article-style with readability
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com/article" --with-readability true --output-format markdown

# Only the element matching a CSS selector
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com" --target-selector "article.main" --output-format markdown

# With cookies and custom user-agent
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com/private" --cookie "session_id=abc123" --with-readability true --json

# Full JSON response
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com" --output-format text --json

Option B: Call API with curl

curl -X POST "https://openapi.felo.ai/v2/web/extract" \
  -H "Authorization: Bearer $FELO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com", "output_format": "markdown", "with_readability": true}'

API Reference (summary)

• Endpoint: POST /v2/web/extract
• Base URL: https://openapi.felo.ai. Override with FELO_API_BASE env if needed.
• Auth: Authorization: Bearer YOUR_API_KEY

Request body (JSON)

Response

Success (200):

{
  "code": 0,
  "message": "success",
  "data": {
    "content": { ... }
  }
}

Fetched content is in data.content; structure depends on output_format.

Error codes

Output Format

• Default output is extracted content only (for direct use or piping).
• If response content is not a string, script prints JSON.
• Use --json when user needs metadata and full response object.

Language

• 預設輸出語言為繁體中文（zh-TW）。除非使用者明確指定其他語言，否則所有摘要、整理、說明等 AI 產出的文字，一律使用繁體中文輸出。
• 網頁原始內容（如簡體中文頁面）維持原文擷取，但 AI 的整理、摘要與說明部分仍須使用繁體中文。

Error response format:

## Web Fetch Failed
- Message: <error message>
- Suggested Action: verify URL/parameters and retry

Important Notes

• ⚠️ 路徑安全：skill 腳本位於 repo root 的 .agents/skills/ 下。若 cwd 不在 repo root，先執行 git rev-parse --show-toplevel 取得路徑再 cd。禁止用 $(...) 語法。
• ⚠️ 指令安全：Copilot CLI 的安全過濾器會封鎖含有特定 shell 語法的指令，執行時禁止加入以下任何寫法，否則指令會被直接擋下：
  • set -e、set -u、set -o、set -euo pipefail 等 set 指令
  • ${var@P}、${!var} 等參數展開語法
  • 巢狀的 $(...) 指令替換
  • 正確寫法：直接執行 cd 與 node 指令，不要包在 shell 腳本控制流程內
• Always require URL before running.
• Validate enum values:
  • output_format: html, markdown, text
  • crawl_mode: fast, fine
• Use --target-selector when users only want a specific part of the page.
• Use --request-timeout-ms for page rendering/extraction wait, and --timeout for local HTTP timeout.
• For long articles or slow sites, consider increasing --timeout.
• API may cache results; use --with-cache false only when fresh content is required.

References

• Web Extract API
• Felo Open Platform