Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Implement

ASecurity

根據 plan package 的 `tasks.md`,以 don't stop until deliver / One-Shot 執行已解鎖 task。一輪恰好 1 個 task 或 Parallel Hint 批次;驗證後立刻回寫 `[X]` 再繼續,直到交付。不能跳步驟。

45 stars
0 votes
0 copies
0 views
Added 9/22/2026
developmentgit

Security Analysis

A100/100

Scanned 9/24/2026

Install to Claude Code

$npx -y skills add Charles5277/nuxt-supabase-starter --skill implement --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Implement?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Implement
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/charles5277-implement/badge)](https://www.skillsdirectory.com/skills/charles5277-implement)

More formats (shields.io, HTML) on the badges page.

Download with Pro
Files
SKILL.md
---
name: implement
description: >-
  根據 plan package 的 `tasks.md`,以 don't stop until deliver / One-Shot 執行已解鎖
  task。一輪恰好 1 個 task 或 Parallel Hint 批次;驗證後立刻回寫 `[X]` 再繼續,直到交付。不能跳步驟。
metadata:
  clade:
    invocation: explicit
disable-model-invocation: true
---
<!-- LOCKED: mirrored from Waterball-Software-Academy/aixbdd@bc8fdebe2d30db384f77911aaed81888ad03bf5f via scripts/sync-upstream-mirrors.ts — edit upstream, never here -->
<!-- clade-skill-scope: both -->

# Implement

`implement` 將 `tasks.md` 視為唯一執行控制平面,但執行上下文必須包含 plan package 與 `truth-delta.md`。它負責交付 plan,不負責重新定義 truth;若發現 truth 與 tasks 明顯矛盾,應停止受影響 task 並回報需要上游修正。

## Operating Principles

- by default 就是 don't stop until deliver / One-Shot:從第一個已解鎖 task 做到目標 plan 全部 `[X]`。使用者不必再寫這句。
- One-Shot 的一輪任務集恰好是 1 個已解鎖 task;有 `Parallel Hint` 時改為 Hint 列出的 `[P]` 批次。One-Shot 不是把後面的序列 task 一次攤開。
- 做完一個任務集就驗證,立刻回寫 `tasks.md` 的 `[X]`,再重算下一筆。未回寫不得開下一個任務集。
- 嚴格執行不能跳步驟。`don't stop until deliver` 不是跳步驟的許可。
- Feature phase 仍序列。`[BDD-GREEN]`、`[BDD-REFACTOR]` 才委派 `/bdd`,並把該 phase 的 `Test Scope` 當範疇。
- Phase 3 的 `[BDD-ALIGN]`、`[BDD-REMOVE]`、`[BDD-RED]` 不委派 `/bdd`;由 subagent 讀 `dsl.md` 該列的 `StepDef 實作語意` 寫測試層。
- Phase 3 review 通過前,不得進入 Feature Green。
- 每一個 task 或 `Parallel Hint` 批次中的每一個 task,在回寫 `[X]` 前都必須各自執行 consumer 的 canonical doctor;doctor 非零不得完成該 task。
- `[BDD-RED]` 的預期 assertion/產品行為失敗不等同 canonical doctor 失敗;保留紅燈測試與證據,並為該 task 另外保存 canonical doctor exit 0 receipt,不得排除測試或以固定成功 wrapper 代替 doctor。
- 全部 tasks `[X]` 後詢問是否 git commit;不得自動 commit。

# SOP

## Phase 1 -- 對齊 plan 並啟動 One-Shot

1. READ 讀取使用者需求、目標 plan package、`tasks.md`、`truth-delta.md` 與目前 task 完成狀態,確認指定範圍(若有)與 One-Shot 起點。
   - 命中 clade lifecycle repo 判準(本次 plan package 的 `plan.md` frontmatter 同時含 `work_id:` 與 `truth_baseline:`)時,`plan.md` 是 lifecycle 檔:系統分析內容改讀同 package 的 `system-analysis.md`,本輪 truth delta 改讀 `plan.md` 的 `## Truth delta` 表,該 package 沒有 `truth-delta.md`。判準只看那兩個 frontmatter 鍵,NEVER 用 repo 名、manifest 或其他檔案存在與否推斷。
2. READ 讀取 `rules/任務選取與連續續跑判準.md`,確認已解鎖 task 的選取、One-Shot 與停止條件。
3. READ 讀取 `rules/TruthDelta測試對齊與BDD委派判準.md`,確認 ADD / MODIFY / DELETE 的執行策略,以及哪些 marker 才委派 `/bdd`。
4. THINK 依已讀任務與判準收斂 One-Shot 起點、是否命中 `Parallel Hint`、所屬 phase;預設做到交付,不停在第一筆。

## Phase 2 -- 補強 repo hygiene 與執行前置

1. READ 讀取 `rules/repo-hygiene-與-ignore-補強判準.md`,確認本次需要檢查哪些 ignore surfaces、哪些只可增補不可覆寫。
2. THINK 依目前 repo、工具鏈、plan package、truth-delta 與 `specs/truth/techstack.md` 收斂本輪必做的最小 hygiene 補強。
3. WRITE 只在實際相關且確有缺口時建立或增補 ignore 設定,避免擴張到無關工具鏈。

## Phase 3 -- 收斂本輪任務集(不能跳步驟)

1. READ 讀取 `rules/嚴格禁止跳步驟判準.md` 與 `rules/平行執行與檔案衝突判準.md`,確認本輪任務集邊界。
2. READ 若 `tasks.md` 當前 phase 含 `Parallel Hint`,讀取 `rules/ParallelHint平行Subagent與衝突Merge判準.md`。
3. THINK 收斂本輪任務集:有 `Parallel Hint` 則為 Hint 列出的 `[P]` 批次;否則恰好 1 個已解鎖 task。
4. THINK 收斂執行模式:Setup 與 Foundational 直接實作並停在該則「只做/不做」;Phase 3 的 ALIGN / REMOVE / RED 走 subagent 測試層;review 走 review 迴圈;`[BDD-GREEN]` / `[BDD-REFACTOR]` 委派 `/bdd`;`[CODE-REMOVE]`、`[REGRESSION]` 直接實作或直接驗證。

## Phase 4 -- 載入精確參照並執行

1. READ 讀取 `rules/技術參照載入與最小上下文判準.md`,確認 `Read`、phase-level `Shared Must Read`、`Extra Read`、truth-delta rows 與相鄰 task / artifact 的載入邊界。
2. READ 依當前任務集的精確參照載入必要 plan artifacts、truth feature、同模組 DSL、實際列出的介面根共用 DSL rows 與既有自動化測試落點;不得掃描或預讀其他模組 DSL。
3. READ 若本輪修改受憲法治理的 artifact 或與 truth specs 發生衝突,讀取 `.agents/constitution/CONSTITUTION.md`、`.agents/constitution/shared.md` 與對應憲法檔。
4. READ 讀取 `rules/自主決策與阻塞處理判準.md`,確認規格缺口、局部矛盾或環境阻礙時的決策優先序。
5. DELEGATE 若本輪是 `Parallel Hint` 批次,為每個 `[P]` task 派出獨立 subagent,prompt 指向該 plan `tasks.md` 的對應任務;全部回來後執行 review 迴圈。
6. DELEGATE 若本輪是 `[BDD-GREEN]` 或 `[BDD-REFACTOR]`,呼叫 `/bdd` 的單一 requested step,傳入 `Test Scope`、action、truth rows、模組 DSL 與實際使用的共用 DSL rows。
7. WRITE 若 task 不帶上述 marker,依 task marker 完成本輪所需的程式、設定、文件或測試變更;不得越出本輪任務與 phase boundary。

## Phase 5 -- 驗證、回寫、繼續 One-Shot 直到交付

1. READ 讀取 `rules/完成定義-驗證與回寫判準.md`,確認 task 完成條件與回寫條件。
2. THINK 依 task 類型收斂最直接的驗證方式。
3. RUN canonical doctor separately for each completed task.
   <!-- clade-doctor-hook: implement-before-task-complete -->
   For a `Parallel Hint` batch, each member gets its own receipt before any member is marked complete. Resolve the command from the consumer's confirmed package manager, workspace root, tech stack, and canonical project check; do not infer `pnpm` or substitute a fixed-success wrapper. A `[BDD-RED]` task may retain its expected assertion/product failure while its separate canonical doctor receipt must exit 0.
4. If doctor is missing, execute the named unlocked Setup bootstrap task first; do not invent a fixed-success wrapper. If doctor is nonzero, preserve the real failure, leave the affected task unchecked, and either fix within the repair task boundary or report the required owner/dependency. An unavailable environment remains blocked with its command, cwd, exit status, and evidence.
5. WRITE only after implementation, direct validation, and that task's doctor receipt all pass: change the corresponding task to `[X]`, preserving every other task state.
6. THINK re-calculate whether an unlocked in-scope task remains; if so, return to Phase 3. This is the One-Shot default.
7. WRITE when the plan package is fully checked and every task has its own doctor evidence, report delivery and ask whether to commit; do not auto-commit. The existing commit 0-C step remains unchanged and receives its own doctor receipt rather than borrowing another task's receipt.

Attribution

YuDefineYuDefine
View sourceMore from Charles5277 →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Browser Extension Developer

Use this skill when developing or maintaining browser extension code in the `browser/` directory, including Chrome/Firefox/Edge compatibility, content scripts, background scripts, or i18n updates.

284722 votes

Seo Optimizer

SEO optimization with keyword analysis, readability assessment, technical validation, content quality. Use for search rankings, blog posts, content audits, or encountering keyword density, readability scores, meta tags, schema markup errors.

2192 votes

Google Official Seo Guide

Official Google SEO guide covering search optimization, best practices, Search Console, crawling, indexing, and improving website search visibility based on official Google documentation

1862 votes

Tanstack Start

Build a full-stack TanStack Start app on Cloudflare Workers from scratch — SSR, file-based routing, server functions, D1+Drizzle, better-auth, Tailwind v4+shadcn/ui. Use whenever the user mentions TanStack Start, asks to scaffold a full-stack Cloudflare app with SSR, wants an SSR dashboard, or asks for a React 19 + Cloudflare Workers app with file-based routing and server functions — even if they don't name TanStack Start specifically. No template repo — Claude generates every file fresh per ...

9881 votes

Pentest

PTES-aligned adversarial security audit for backend, frontend, and mobile applications. Produces a CVSS-scored Hacker Report with verified PoCs and phased remediation.

5491 votes
View all in development →