Codex 統測考古題下載與整理練習

練習目標

Module 3 全部練習預計 30~40 分鐘完成,本頁是第一部分:練習用一段自然語言指令,請 Codex 到官方網站下載統測歷屆考古題,並依年份整理成資料夾。不需要先設計複雜的分步驟 prompt,直接告訴 Codex 你要什麼結果就好。完成本頁後,可以接著做第二部分:Codex 統測試題切題與答案鍵練習,把下載好的試題 PDF 切成一題一張圖片,並建立 JSON 格式答案鍵。

完成後你應該能夠:

練習範圍與資料夾結構

本次練習以下面的範圍為例,你也可以換成別的科目群:

項目內容
考試四技二專統一入學測驗(統測)
科目群電機與電子群資電類
學年度111、112、113、114、115(共 5 個學年度)
要下載的檔案專業科目(一)試題、專業科目(二)試題、專業科目(一)答案、專業科目(二)答案(每個學年度 4 個檔案)

下載完成後,請 Codex 依照學年度建立資料夾,例如:

111/
  專業一題目.pdf
  專業二題目.pdf
  專業一答案.pdf
  專業二答案.pdf
112/ ... 115/(同樣四個檔案)

這些試題與答案是技專校院入學測驗中心官網公開、免登入提供的考古題資料,練習時不需要繞過任何登入或驗證碼。

參考 Prompt

網站操作版

如果想寫得像傳統爬蟲腳本一樣精確,甚至可以連網址、要點哪一個連結都寫死。這種寫法在網站介面沒變的時候最可靠,但只要官網改版、連結文字換了,整段 Prompt 就可能直接卡住。

請依照以下步驟操作,下載 111 到 115 學年度統測「電機與電子群資電類」的專業科目(一)(二)試題與標準答案:

1. 開啟 https://www.tcte.edu.tw/index.php?mod=TVETest%2Fdown_exam4y
   這是「技專校院入學測驗中心:四技二專試題下載」列表頁,裡面是一個依學年度排列的表格。

2. 對於每個目標學年度(111、112、113、114、115),在表格中找到「年度」欄等於該學年度的那一列,
   點擊該列「試題答案及組距」欄位的「查閱」連結。
   這會開啟 https://web1.tcte.edu.tw/EXAM/{學年度}_4y/ 這個頁面(例如 115 學年度是
   https://web1.tcte.edu.tw/EXAM/115_4y/)。

3. 在該學年度頁面中,找到「04 電機與電子群資電類」這個科目群的區塊,裡面會列出:
   - 專業科目(一)試題
   - 專業科目(一)標準答案
   - 專業科目(二)試題
   - 專業科目(二)標準答案
   每個項目通常同時提供 PDF 和 Word(docx)兩種格式;如果兩種都有,請下載 PDF,
   因為官網有註明「請以PDF檔案為準」。只有在沒有 PDF 時才下載 Word 檔。

4. 下載完成後,依照學年度整理成資料夾,資料夾名稱用學年度數字(111、112、113、114、115),
   每個資料夾裡放該學年度的「專業一題目」「專業二題目」「專業一答案」「專業二答案」四個檔案。

5. 如果某個學年度的頁面版面跟上面描述的不一樣(例如連結文字換了、找不到「查閱」連結、
   科目群代碼不是 04),請停下來、描述你實際看到的畫面,不要用猜的硬點下去,也不要跳過。

詳細版

先看一個把細節都寫清楚的版本:說明要去哪裡找、限定科目群與學年度範圍、逐一列出每年要下載的 4 個檔案、交代資料夾與檔名規則,並要求 Codex 遇到缺漏要主動回報、完成後給一份檢查清單。

請到「技專校院入學測驗中心」官方網站,找到四技二專統一入學測驗(統測)的歷屆試題/考古題下載頁面。

目標科目群:電機與電子群資電類
目標學年度:111、112、113、114、115 學年度(共 5 個學年度)

請依序處理每個學年度,下載以下 4 個檔案:
1. 專業科目(一)試題
2. 專業科目(二)試題
3. 專業科目(一)標準答案
4. 專業科目(二)標準答案
其他科目(例如共同科目、術科、其他群別)不用下載。

下載完成後,請依照學年度整理成資料夾,資料夾名稱直接用學年度數字(111、112、113、114、115),
每個資料夾裡放該學年度的「專業一題目」「專業二題目」「專業一答案」「專業二答案」四個檔案,
檔名請清楚標示是題目還是答案、是專業一還是專業二。

處理過程請注意:
- 這些資料是官網公開、免登入即可下載的考古題,不需要登入帳號、繞過驗證碼,也不要嘗試存取任何需要權限的頁面。
- 如果某個學年度在官網上找不到對應檔案(例如尚未公告或格式不同),請明確告訴我是哪個學年度、哪個檔案缺漏,不要略過不提。
- 全部下載整理完成後,請列出一份檢查清單,逐一確認 5 個學年度、20 個檔案是否都已到位。

簡易版

同一個任務也可以只講你要的結果,把「怎麼做」留給 Codex 自己判斷:

請到技專校院入學測驗中心的網站,幫我下載 111 到 115 學年度統測「電機與電子群資電類」的資料。
我只需要專業科目(一)、專業科目(二)的試題,還有這兩科的標準答案。
下載好之後,依照學年度整理,一個學年度一個資料夾,資料夾名稱用 111、112 這樣,
裡面放專業一題目、專業二題目、專業一答案、專業二答案。

三種寫法怎麼選? 「網站操作版」把網址、要點哪個連結都寫死,這就是傳統爬蟲腳本的寫法——網站介面不變時最穩定,但官網一旦改版、連結文字或版面換了,整段 Prompt 可能直接卡住,因為它被綁死在你當初看到的那個畫面上。「詳細版」只講規則和驗收標準(科目群、學年度範圍、要幾個檔案、遇到缺漏怎麼辦),不綁定網站的操作細節,Codex 可以自己判斷去哪裡找、按鈕文字換了也能自己摸索。「簡易版」連規則都不講,只講你要的結果,把「怎麼做」完全留給 Codex——它有最大的空間自己摸索網站介面、嘗試不同做法,有時候反而能找到更好的路徑,但驗收標準也最模糊,比較難確認做得夠不夠完整。多數時候,介於詳細版和簡易版之間最實用:講清楚你要什麼結果、什麼算做完,但不要規定 Codex 該怎麼操作網站。

檢查重點

反思問題

  1. 如果只用瀏覽器手動下載,這 5 個學年度、20 個檔案要花多少時間?跟請 Codex 做比起來差在哪裡?
  2. 如果要換成別的科目群(例如機械群),這段 Prompt 只需要改哪一個詞?
  3. 詳細版把步驟都寫清楚,一定比簡易版可靠嗎?如果詳細版寫的步驟跟官網實際介面不符,會發生什麼事?
  4. 如果技專校院入學測驗中心明年改版,把「查閱」改成別的字、頁面版面也不一樣了,網站操作版的 Prompt 會發生什麼事?簡易版又會發生什麼事?

Codex Exam Archive Download and Organize Practice

Practice Goal

Module 3 is designed to take about 30–40 minutes in total. This page covers Part 1: practicing how to give Codex one natural-language instruction to download past exam papers from the official site and organize them into folders by year. No need to design a complex multi-step prompt first — just tell Codex the result you want. Once you've finished this page, continue with Part 2: Codex Exam Question Cropping and Answer Key Practice, which crops the downloaded exam PDFs into one image per question and builds a JSON-style answer key.

After this, you should be able to:

Scope and Folder Layout

This practice uses the following scope as an example; swap in a different subject group as needed:

ItemValue
ExamFour-Year Technology and Two-Year Program Joint College Entrance Examination (統測)
Subject groupElectrical and Electronics Group — Information/Electronics Track (電機與電子群資電類)
Academic years111, 112, 113, 114, 115 (5 years)
Files to downloadProfessional Subject (1) exam, Professional Subject (2) exam, Professional Subject (1) answer key, Professional Subject (2) answer key (4 files per year)

Once downloaded, have Codex create one folder per academic year, for example:

111/
  專業一題目.pdf
  專業二題目.pdf
  專業一答案.pdf
  專業二答案.pdf
112/ ... 115/ (same four files each)

These exam papers and answer keys are public archival materials that the official site already provides for free, with no login required — there is nothing to bypass.

Reference Prompt

Site Navigation Version

You can also write it as precisely as an old-school scraper script: hard-code the URL and exactly which link to click. This is the most reliable version as long as the site's layout doesn't change — but the moment the official site gets redesigned or a link's wording changes, the whole prompt can get stuck.

Follow these steps to download the Professional Subject (1) and (2) exam papers and standard answer keys for
the "電機與電子群資電類" (Electrical and Electronics Group — Information/Electronics Track) for the 111 to 115
academic years:

1. Open https://www.tcte.edu.tw/index.php?mod=TVETest%2Fdown_exam4y
   This is the TCTE "Four-Year Technology exam downloads" listing page: a table of academic years.

2. For each target academic year (111, 112, 113, 114, 115), find the table row where the "年度" (Academic
   Year) column equals that year, then click the "查閱" (view) link in the "試題答案及組距" (Exam Questions,
   Answers & Score Ranges) column.
   This opens https://web1.tcte.edu.tw/EXAM/{year}_4y/ (for example, the 115 academic year page is
   https://web1.tcte.edu.tw/EXAM/115_4y/).

3. On that year's page, find the section for "04 電機與電子群資電類" (subject group 04) and locate:
   - Professional Subject (1) exam paper (專業科目(一)試題)
   - Professional Subject (1) standard answer key (專業科目(一)標準答案)
   - Professional Subject (2) exam paper (專業科目(二)試題)
   - Professional Subject (2) standard answer key (專業科目(二)標準答案)
   Each item is usually offered in both PDF and Word (docx). If both are available, download the PDF — the
   site itself notes "請以PDF檔案為準" (PDF is authoritative). Only use the Word file if no PDF exists.

4. Once downloaded, organize the files into one folder per academic year, named with the year number alone
   (111, 112, 113, 114, 115), each holding that year's Subject 1 exam, Subject 2 exam, Subject 1 answer key,
   and Subject 2 answer key.

5. If any year's page layout doesn't match this description (different link wording, no "查閱" link, subject
   group not coded 04), stop and describe exactly what you see instead of guessing where to click — don't
   skip it.

Detailed Version

Here's a version that spells out the details up front: state where to look, scope the subject group and academic-year range, list exactly which 4 files to download per year, spell out the folder and filename rules, and ask Codex to actively report any gaps and give a checklist when done.

Go to the official website of the Technical and Vocational Education Joint College Entrance Examination Center,
and find the past-exam-paper download page for the Four-Year Technology and Two-Year Program Joint College
Entrance Examination (統測).

Target subject group: Electrical and Electronics Group — Information/Electronics Track (電機與電子群資電類)
Target academic years: 111, 112, 113, 114, 115 (5 years total)

For each academic year, download these 4 files:
1. Professional Subject (1) exam paper
2. Professional Subject (2) exam paper
3. Professional Subject (1) standard answer key
4. Professional Subject (2) standard answer key
Skip everything else (common subjects, practical/skills tests, other subject groups).

Once downloaded, organize the files into one folder per academic year, named with the year number alone
(111, 112, 113, 114, 115). Each folder should contain that year's Subject 1 exam, Subject 2 exam, Subject 1
answer key, and Subject 2 answer key, with filenames that clearly indicate exam vs. answer key and Subject 1
vs. Subject 2.

While doing this:
- These files are public archival materials available on the official site with no login required — do not
  log in, bypass a CAPTCHA, or try to access any restricted page.
- If a file can't be found for a given year (e.g. not yet published, or in a different format), tell me
  explicitly which year and which file is missing — don't skip it silently.
- When everything is downloaded and organized, give me a checklist confirming all 5 years and all 20 files
  are accounted for.

Simple Version

The same task can also be stated as just the result you want, leaving "how" up to Codex:

Go to the Technical and Vocational Education Joint College Entrance Examination Center website and download
the 111 to 115 academic year exam materials for "電機與電子群資電類" (Electrical and Electronics Group — Information/Electronics Track).
I just need the Professional Subject (1) and Professional Subject (2) exam papers, plus the standard answer
keys for both subjects.
Once downloaded, organize everything by academic year: one folder per year, named 111, 112, and so on, each
containing the Subject 1 exam, Subject 2 exam, Subject 1 answer key, and Subject 2 answer key.

Which of the three should you use? The site navigation version hard-codes the URL and exactly which link to click — that's literally how an old scraping script is written. It's the most reliable while the site's layout stays the same, but the moment the official site changes even a link's wording, the whole prompt can get stuck, because it's pinned to the exact screen you saw when you wrote it. The detailed version only states the rules and acceptance criteria (subject group, year range, how many files, what to do about gaps) without pinning down the site's click path, so Codex can figure out where to click on its own, even if button text changes. The simple version doesn't even state the rules — just the result you want — leaving "how" entirely to Codex; it has the most room to adapt, and sometimes finds a better path or catches an edge case you hadn't thought of, but the acceptance criteria are the fuzziest, so it's harder to tell whether it did a thorough job. Most of the time, the sweet spot sits between the detailed and simple versions: state clearly what result you want and how you'll know it's done, but don't dictate how Codex should operate the website.

Check Points

Reflection Questions

  1. If you downloaded these 5 years' worth of files (20 files total) manually in a browser, how long would it take — and what's different about asking Codex instead?
  2. If you switched this to a different subject group (for example, the Mechanical group), which single word in the prompt would you need to change?
  3. Does spelling out every step in the detailed version always make it more reliable? What happens if the steps it specifies don't match the site's actual layout?
  4. If TCTE redesigns the site next year — renaming the "查閱" link, changing the layout — what happens to the site navigation version of the prompt? What happens to the simple version?