16 關免費主線已開放 · Skill 池內測期免費領取 從第一關開始
菜鸟skill
菜鸟skill
新手學AI
部署網站一條龍自媒體創作一條龍職場提效一條龍商業變現一條龍
Skills 定價
語言:
English 繁體中文 日本語 Español
真實產出 · 已記錄檢查 真實樣例 · Markdown
30 秒看完完整流程

Three small tasks, three saved results

Inspect a folder

first-agent observed two visible files and saved a directory report. Its result was subsequently rendered in the preview and video.

Make a precise edit

first-task created a three-line document and changed only the second line. Its saved before/after documents were checked and used in the source-linked walkthrough.

Use an installed Skill

The complete weekly-report-studio ZIP was installed in an isolated directory. Its installed rules were read and applied to current production notes; its validator accepted the new report. That report was then consumed by the installation walkthrough.

These are actual local production operations on public practice/development material. No email, post or customer publication occurred. The journal checker verifies saved file evidence, not human comprehension or universal Agent compatibility.

樣例內容原樣呈現，僅做閱讀排版 下載完整樣例
運行方式 Codex 由你的 Agent 執行，模型用量由服務商計費
實測環境：
Codex · Windows
Agent／模型費用依供應商而定；額外服務與依賴詳見完整包。
Two useful Skills; Node.js 18+ for checks

主線哪幾關用到它

Skill 是一把武器，配合新手主線實操，打通從「會玩」到「精通」的打怪升級全流程。

1-4 · 用 skill 解决三件真实小事 →

帶著這次成果繼續: 把需求说清楚：验收標準先行 →

✓ 適合誰 (Good Fit)
  • 沿主線動手學習的 AI 新手
  • 從對應關卡開始；頁面瀏覽不會被當成任務完成。
⊗ 不適合誰 (Not For)
  • 超出本次任務的無人值守操作或未核准的帳號存取
✦ 你會得到 (What You Get)
  • 建立「這活能不能让 agent 干」的直觉，完成首批真实产出
  • 两件小事的产出被你实际用过（发出去了/交上去了）
  • 至少一件用到了真实素材而不是演示数据

交付內容：完整 SKILL.md 規則文件 + 公開自測樣例。需要在你自己的 AI Agent 中運行。

怎麼開始？

Skill 是一份教 AI 完成任務的規則。

領取並安裝 ↑
1
提供手上的材料
挑本站两个 skill，各完成一件你自己的真实小事（不是演示任务）。
2
讓 Agent 按規則完成
建立「這活能不能让 agent 干」的直觉，完成首批真实产出
3
打開並核對真實產出
两件小事的产出被你实际用过（发出去了/交上去了）
發給你的 AI Agent

用 daily-three 帶我完成這一關。缺少材料先問我，實際使用工具並儲存結果，再按關卡判據核對。

使用前了解 證據會標明實測範圍；模型與權限依你的 Agent 而異。解壓成功不等於已被 Agent 自動發現。 Agent／模型費用依供應商而定；額外服務與依賴詳見完整包。

從這裡開始使用

压缩包里是规则、参考库、脚本、真实产出和演示片，不只是一个提示词文件。

下載完整包（.zip） 下載 SKILL.md
  1. 打開 Codex 或 Claude Code，粘貼安裝提示詞。
  2. 讓 Agent 讀取並保存完整規則，確認安裝成功。
  3. 複製上方使用示例，換成你的素材開始運行。
手動安裝 · 查看完整 SKILL.md（90 行）

Codex：~/.agents/skills/daily-three/SKILL.md
Claude Code：~/.claude/skills/daily-three/SKILL.md

--- name: daily-three description: Help a beginner finish two or three genuinely useful small tasks using installed Skills, inspect each saved result and keep an evidence-backed task journal without pretending files were published or used by the learner. --- # Use Skills on real small tasks This is level 1-4 of the newbieskill main quest. Move from trying an example to choosing tasks the learner actually needs to finish. ## Result - Two or three usable saved results from at least two different Skills. - A short task journal connecting the original need, chosen Skill, result and checks. - A clear next task based on what the learner just made. ## Choose small, real tasks Ask what the learner needs to finish today if the conversation contains no concrete needs. Offer a short menu grounded in installed capabilities, such as checking a project folder, revising a note or drafting a weekly report. Let the learner choose; do not turn a small task into a complex workbench. Prefer a result they can inspect and use immediately. Start with two tasks. Add a third only when the learner has a useful reason. Public examples demonstrate the method; they do not substitute for the learner's own material. ## Check the selected Skills Read the actual installed rules and identify their required inputs and dependencies. Check whether their supporting files are present. Do not infer that a catalog page or a saved screenshot is a complete, tested installation. If a dependency needs another account or fee, explain it before running that task. Choose an already available alternative when it serves the learner's need. ## Execute one task at a time 1. State a small result and one or two checks in the learner's words. 2. Collect the actual input or ask one focused clarification. 3. Invoke the selected installed Skill and use real tools where the task requires them. 4. Save a named output; keep the original input and existing files intact. 5. Open or read the saved result. Check facts, content and usability against the input. 6. Record what passed, what needs another step and whether the learner has actually used the result. Do not say the second task ran merely because the first task completed. If one task fails, keep the successful result and identify the failed task's blocker. ## Keep a useful journal For each task record `title`, `skill`, `output` and a short `check` explanation. Use output paths relative to an explicitly selected base directory. Set `used` to `true` only after the learner confirms use, or when an actual local follow-on operation consumed the result. Record that use in `usedFor`; “file exists” is not enough. The read-only helper checks the journal's files and produces hashes: ```text node scripts/check-journal.mjs <journal.json> <selected-base-directory> ``` The helper verifies distinct tasks, distinct saved files, at least two Skills, file existence and a use/check record. It cannot establish the truth of a learner's claim or the quality of a document by hashing it. The Agent must still inspect the content and explain the result. ## Handle missing or incomplete material - No real task: ask what the learner needs to hand in, send or save today. - No inputs: request the minimum missing material, not a biography or a complete project archive. - One complete task only: save it and help finish a second useful task. - Result not used yet: say “ready to review/use”; do not mark actual use as verified. - Sending/publishing requires separate authorization; local creation is not permission to send it. ## Self-test tasks 1. Trace three actual local production results to their source Skills and their real follow-on uses; verify the saved file hashes. 2. Build a different two-task journal and check that the minimum useful session passes without inventing a third task. 3. Submit an empty journal and a journal naming a nonexistent output; both must fail without creating those outputs. Any fabricated execution, unobserved publication, missing file or repeated file counted as two tasks fails acceptance. An artifact check is not a new Agent run. Record the scope in the evidence. ## Continue the main quest Return: `/path/daily-three/`. Next: `/path/prompt-spec/` — give your next task clear acceptance criteria before running it. Branches can include reports, notes, presentations and publishing formats when the learner's actual task needs them. ## Rights Original newbieskill workflow. Personal and commercial work is permitted; do not resell the complete package.

新手常問 (FAQ)

有疑惑？先看看這裡，買前明明白白，用時順順利利。

怎麼開始？ +
下載完整 ZIP，把安裝提示交給 Agent，保留腳本與參考檔案。模型費用另計。
主線 Skill 免費嗎？ +
這些基礎主線 Skill 免費；進階工作室商品沿用各自定價。
菜鸟skill 會員

讓每一種重複，
都有解法。

一個訂閱，解鎖全部工作室 Skills。只需要這一款？也可以單獨獲取。

比較完整方案
免費 用 skill 解决三件真实小事
免費獲取