跳轉到

Subagent Cookbook — 15 個複製貼上就能用的派遣 recipe

繁體中文 | 简体中文 | English

📋 這是什麼Stage 5.5 教你 subagent 是什麼概念、這份 cookbook 教你今天就能用——15 個情境、每個含「該用哪個 subagent + 完整 prompt 範本(複製即可用)+ 何時不用」。

⚠️ 第一次看?先讀 Stage 5.5 的「可派遣的 subagent 有哪些」+「decision table」兩節——理解「什麼是 subagent」「Claude Code 內建有哪些」之後再來查 recipe。


怎麼讀這份 cookbook

每個 recipe 用同樣的 4 段結構:

內容 為什麼有這段
情境 你今天工作會遇到的具體場景 從「我有 X 問題」找到 recipe、不從「我想用 subagent」找
Subagent 用哪一個(Claude Code 內建名稱) 直接 copy 名字、不用想
Prompt 範本 複製貼上即可用的指令文字 不用自己想怎麼寫
何時不用 比用 subagent 更好的替代方案 避免「殺雞用牛刀」

💡 怎麼實際派遣 subagent:在你的 Claude Code 終端機對話框裡、直接輸入(或貼上)prompt 範本——就這樣。Claude 看到指令、會自動透過 Task tool(內部派遣機制)找到對應 subagent 跑、跑完回主 session 一段摘要。不需要 slash command、不需要特殊語法

📌 subagent ≠ slash command/agents 是查當前可用 subagent 的指令、不是用來「呼叫」subagent。派遣 subagent 直接打對話 prompt 文字即可。完整對比表(subagent vs skill / vs slash command / description router)見 Stage 5.5 §易混淆觀念釐清


先確認你有哪些 subagent 可用

在你的 Claude Code session 內跑 /agents 一個指令、會列出全部當前可用的 subagent(內建 + plugin + 自訂)。

Claude Code 預設內建 7 個 subagent:general-purpose / code-reviewer / Explore / Plan / frontend-developer / claude-code-guide / statusline-setup(截至 2025-11、可能會變動)。

📌 每個內建 subagent 的功能說明 + 「遇到 X 任務用 Y」decision table 的 canonical 在 Stage 5.5 §可派遣的 subagent 有哪些——本 cookbook 聚焦「怎麼派遣」的 15 個 recipe,內建清單與選用邏輯以 Stage 5.5 為準。

💡 如果 /agents 列表跟這份 cookbook 不一致:表示你裝了 plugin 或 Claude Code 版本不同。recipe 名字對不上時、找最接近的 subagent 用就好(例如沒有 Explore、用 general-purpose 也能跑搜尋)。

確認完當前可用清單之後、就可以照下面 15 個 recipe 派遣了——找到符合你情境的 recipe、複製 Prompt 範本貼進 Claude Code 對話框即可


15 個 Recipe

Recipe 1: 寫完一批新 code、要 commit 前的 review

情境:你剛寫了 ≥ 50 行新 code、想 commit 但怕有 bug / security 問題沒看到

Subagentcode-reviewer

Prompt 範本

Review the staged changes in this repo (git diff --cached).
Focus on: (1) security issues (hardcoded secrets, SQL injection, XSS),
(2) error handling gaps, (3) missing tests, (4) violations of CLAUDE.md
conventions. Per-category PASS/FAIL + concrete fix for each issue with
file:line reference + overall verdict (APPROVE / REQUEST CHANGES).

何時不用:< 20 行的 typo / formatting fix(直接 git diff 自己掃就好、不用花 subagent token)


Recipe 2: 進新 repo、不知該從哪個 file 開始讀

情境:clone 完一個別人的 repo、README.md 講不清楚程式入口在哪、不想隨便挖

SubagentExplore

Prompt 範本

Map the entry points and core structure of this codebase. Report:
(1) main entry script(s) — what file gets run first,
(2) core module organization — top 3-5 directories and what each does,
(3) test directory layout,
(4) any "where to start reading" guidance in docs/.
Under 300 words with file paths.

何時不用:你已經知道要看的 file 路徑(直接用 Read tool)


Recipe 3: 設計 refactor / migration plan(先想清楚再動手)

情境:要重構一個大 module 或做 framework migration、不確定步驟、想先有 plan 再開始

SubagentPlan

Prompt 範本

Design a step-by-step plan to refactor <module-name> (currently a single
file with tight coupling) into clear, testable components with explicit
interfaces. Include: (1) phased breakdown (≤ 5 phases), (2) which files
touched per phase, (3) what tests gate each phase, (4) rollback strategy
if a phase fails. Don't write code — just the plan.

何時不用:refactor 只動 1-2 個 file(直接動手、不需要 plan overhead)


Recipe 4: 多檔 / 跨 locale parity 審查

情境:你改了 zh-TW + zh-Hans + en 三個 mirror、想確認三邊內容一致

Subagentcode-reviewer

Prompt 範本

Review the staged diff for cross-locale parity across the 3 locale
variants of <file-stem> (.md / .zh-Hans.md / .en.md). Check: (1) same
section structure (## headers match), (2) same table row counts, (3)
same required terms present in each locale, (4) locale conventions
correct (zh-Hans uses "" not 「」; en uses English). Report per-file
PASS/FAIL.

何時不用:只改 1 個 locale(沒 parity 問題)


Recipe 5: 多 source fact-check / research

情境:要寫一段引用、不確定多個 source 講的是不是同一件事、需要交叉驗證

Subagentgeneral-purpose

Prompt 範本

Fact-check this claim: "<claim>". Search for: (1) primary source / official
docs / paper, (2) 2-3 independent secondary sources, (3) any contradictions
or version differences. Report: confirmed / contradicted / nuanced, with
direct quotes and URLs. Under 400 words.

何時不用:你已經有 1 個權威 source(直接讀那一個就好)


Recipe 6: 找某個 symbol / function 在哪定義

情境:codebase 內到處在 import parse_config、想知道實作在哪個 file

SubagentExplore

Prompt 範本

Find where `<symbol-name>` is defined in this codebase. Report: (1) the
file:line where it's defined, (2) what it does (1-line summary), (3) which
files import / use it (top 5). Use Grep, not full file reads.

何時不用:你已經用 IDE 的 "Go to definition" 可以跳過去(IDE 比 subagent 快)


Recipe 7: 比較多 paper / source 的 claim 是否一致

情境:寫文獻回顧、3 篇 paper 對同一件事說法不同、不知該信誰

Subagentgeneral-purpose

Prompt 範本

Find and compare 3 independent sources on the topic <topic>. For each
source, capture: title + author + year + URL. Then report: (1) what
they agree on, (2) where they disagree (with direct quotes), (3) which
is most recent / authoritative, (4) suggested phrasing if you had to
summarize the consensus. Flag if a key source is behind a paywall and
unreadable.

何時不用:3 篇都是同一個作者 / 同 lab(沒「多 perspective」可比)


Recipe 8: Release commit security audit

情境:要發 v1.0 / 升 major version、想最後一次安全掃描

Subagentcode-reviewer

Prompt 範本

Pre-release security audit for <release-tag> (use git log to find the
relevant commit range since last release). Check: (1) hardcoded
credentials / API keys, (2) eval() / exec() / shell injection risks,
(3) deprecated dependencies with known CVEs, (4) any auth / session
handling changes since last release, (5) public API surface changes
(breaking? documented?). Per-category PASS/FAIL + remediation per
finding.

何時不用:patch release(hotfix)只動 1 行(人工掃 < 1 分鐘)


Recipe 9: 評估架構變動的 blast radius

情境:想改一個 base class / 共用 utility、不確定會影響多少 file

SubagentPlan

Prompt 範本

Assess the blast radius of changing <component>. Report: (1) direct
dependents (which files import this), (2) indirect dependents (transitive
imports), (3) tests that gate the change, (4) suggested rollout order
(safest → riskiest), (5) feature flag / kill switch strategy if available.
Don't make the change yet — just the impact analysis.

何時不用:只影響 1 個 file 內部(沒 blast radius 概念)


Recipe 10: Spawn 並行 subagent 跑同一任務 × 多目標

情境:4 個 branches file 都要做同樣的「academic-style audit」、想一次跑 4 個 parallel

Subagentgeneral-purpose × N(同時 spawn 多個)

💡 怎麼「同時 spawn 多個」:在 Claude Code 對話框內、連續輸入下面的 prompt 4 次(每次換 <file-path>)。Claude Code 會自動並行跑、不需要等第一個 subagent 結束才能輸入第二個——這就是「parallel spawn」的實際操作方式。

Prompt 範本(每次換 <file-path>):

Audit `<file-path>` for academic-style issues: (1) over-engineering
jargon without first-use explanation, (2) clarity (long sentences,
vague pronouns), (3) unsupported % claims, (4) persona-fit (wrong
technical level for the file's stated target audience — see the
file's banner / intro callout for who it's for). Report 4-category
PASS/FAIL + fix-list with line numbers. Under 500 words.

何時不用:1 個目標就夠(直接呼叫 1 次、不要 over-orchestrate)


Recipe 11: 找跨 repo 的相似 implementation

情境:你在某個 repo 看到一個 pattern、想知道另一個 repo 有沒有類似實作

SubagentExplore

Prompt 範本

In <repo-path>, search for code patterns similar to: <description of
pattern, e.g., "retry decorator with exponential backoff">. Report:
(1) up to 5 candidates with file:line, (2) brief description of each,
(3) which is most idiomatic for this codebase's style.

何時不用:你有具體 function 名要找(用 Recipe 6 / Explore 加精確 grep 更快)


Recipe 12: LLM-as-judge eval(structured PASS / FAIL)

情境:跑了 100 個 test case、要評每個 output 是否符合 spec、不想人工看

Subagentgeneral-purpose

Prompt 範本

Evaluate the agent outputs in <eval-file.jsonl> against the spec stated
in the file's first 5 lines (or in <spec-file.md>). For each row: PASS /
FAIL with 1-sentence reason. Aggregate: pass rate + top-3 failure modes.
Output as structured JSON: {"case_id": "...", "verdict": "PASS|FAIL",
"reason": "..."}.

何時不用:< 5 個 case(自己看更快);evaluation 涉及主觀判斷(LLM judge 不可靠)


Recipe 13: UI component design / accessibility audit

情境:寫了一個 React component、想確認 ARIA + keyboard nav + responsive 都做對

Subagentfrontend-developer

Prompt 範本

Audit <component-file> for: (1) ARIA roles + labels (screen reader
compatibility), (2) keyboard navigation (tab order, Enter / Esc / arrows
behavior), (3) responsive breakpoints (mobile 360px / tablet 768px /
desktop 1280px), (4) color contrast (WCAG AA), (5) touch target size
(≥ 44px). Report per-category findings + fixes.

何時不用:純後端 / CLI 工具(沒 UI 可 audit)


Recipe 14: 問 Claude Code feature 怎麼用

情境:忘了 hooks 怎麼設、忘了 slash command 的 frontmatter 欄位、想查文件

Subagentclaude-code-guide

Prompt 範本

How do I <specific Claude Code feature question, e.g., "configure a
PreToolUse hook to block dangerous bash commands">? Show: (1) minimum
config in settings.json, (2) example hook script (Python or shell),
(3) 1 common gotcha, (4) where in the official docs to read more.

何時不用:你已經寫過幾次(直接看自己 ~/.claude/settings.json 範例更快)


Recipe 15: 寫 React form validation 邏輯

情境:要做一個 sign-up form、要 email format / password strength / real-time validation

Subagentfrontend-developer

Prompt 範本

Implement a React sign-up form with: (1) email format validation
(real-time, debounce 300ms), (2) password strength meter (≥ 8 chars,
mixed case, digit, symbol), (3) inline error messages with ARIA
live region, (4) submit button disabled until valid. Use <library, e.g.,
React Hook Form + Zod>. Include: component code, validation schema,
1 happy-path test, 1 error-path test.

何時不用:非 React stack(用 Vue / Svelte 對應的 subagent、或 general-purpose


Recipe 索引(按 subagent type 找)

不確定該用哪個 subagent?從任務類型反查:

Subagent Recipes
code-reviewer 1(pre-commit review)/ 4(cross-locale parity)/ 8(release security audit)
Explore 2(new codebase)/ 6(find symbol)/ 11(cross-repo similar code)
Plan 3(refactor plan)/ 9(blast radius)
general-purpose 5(fact-check)/ 7(multi-paper compare)/ 10(parallel multi-target)/ 12(LLM-as-judge eval)
frontend-developer 13(a11y audit)/ 15(React form)
claude-code-guide 14(Claude Code feature 查詢)

何時 NOT 該用 subagent

Subagent 不是免費的——每次派遣燒 token、有延遲。下面 4 種情境不該用 subagent、自己跑更划算:

  1. 任務 < 5 分鐘可自己完成 — 殺雞用牛刀;subagent overhead 不划算
  2. 結果需要逐步 user feedback — subagent 是「派出去、跑完回報一次」、不能中間問你;要逐步確認的任務直接在主 session 跑
  3. 任務需要主 session 的 context memory — subagent 是獨立 context window、看不到主 session 前面的對話;要用「我們剛剛討論的 X...」這種 reference 的任務、不適合
  4. 任務涉及 conversation-level judgment — 像「這個 architecture 決策該怎麼權衡」、需要對話協作、不適合丟給 subagent

💡 判斷的快速辦法:如果任務可以用「寫一份完整 brief 給陌生人接手做」描述清楚 → 適合 subagent;如果需要「我跟你討論一下」→ 留在主 session。


接下來