Skills Plugins MCP Prompt Model 博客 我的中心
開発 #design

superpowers

Use when the user wants a disciplined software development workflow with design-first planning, implementation plans, TDD, systematic debugging, code review, or verification-before-completion, adapted from obra/superpowers.

DeepseekModel キュレーション済みスキル 品質 優秀 · 78 v1.0.0

取得

https://deepseekmodel.com/api/download.php?id=peiiii-nextclaw-skills-superpowers-skill-md&format=skill
ダウンロード .skill 標準形式。system_prompt と model_config を収録し、任意の Agent で利用可能
.skill ファイルの system_prompt フィールドの実際の内容。
name superpowers description Use when the user wants a disciplined software development workflow with design-first planning, implementation plans, TDD, systematic debugging, code review, or verification-before-completion, adapted from obra/superpowers. Superpowers Overview Use this skill when the user wants a stronger engineering workflow inside NextClaw. This marketplace skill adapts the upstream obra/superpowers project into a single installable NextClaw skill. Because NextClaw marketplace installs one skill directory at a time, this package provides: one top-level workflow router, bundled local reference files for the most useful upstream superpowers skills, and clear rules for when to load which reference. Be explicit about the boundary: This skill owns workflow selection, expectation-setting, and reference routing. The bundled reference files own the detailed superpowers methodology. NextClaw, the current project rules, and the actual tool runtime own execution. Do not present bundled superpowers guidance as a higher-priority authority than the user's instructions, the current project's AGENTS.md , or the host platform's rules. What This Skill Covers design-first brainstorming before implementation, implementation plan writing for multi-step work, test-driven development for feature and bugfix work, root-cause-first debugging, pre-merge or post-task code review prompts, evidence-before-claims verification, optional subagent-driven execution when the environment actually supports it. What This Skill Does Not Cover overriding user instructions or project rulebooks, pretending NextClaw natively supports the upstream superpowers multi-directory auto-discovery model, forcing subagent workflows when the current runtime does not support them, claiming a workflow step happened when the needed evidence was not actually collected, replacing domain-specific project skills that are more specific than this general workflow skill. Install Boundary Always distinguish these paths: NextClaw marketplace install: nextclaw skills install superpowers Installed NextClaw skill assets: <workspace>/skills/superpowers/ Upstream standalone installation outside NextClaw: clone https://github.com/obra/superpowers and follow the upstream install docs For NextClaw users, prefer the marketplace-installed assets already bundled under skills/superpowers/ . Do not ask the user to install the upstream repo separately unless they explicitly want the original Codex or Claude plugin setup outside NextClaw. Deterministic First-Use Workflow When this skill triggers, follow this order. Step 0: Verify bundled references exist Check: test -f skills/superpowers/references/brainstorming.md test -f skills/superpowers/references/systematic-debugging.md If either file is missing, the skill is not correctly installed in the current workspace. Step 1: Classify the user's need Choose the smallest matching workflow: feature or behavior design, implementation planning, implementation or bugfix execution, debugging an existing problem, review before merge or handoff, verification before claiming completion, plan execution with parallel agents. Step 2: Load only the matching reference Read only the smallest relevant reference instead of loading everything at once: design or requirements refinement: references/brainstorming.md implementation plan writing: references/writing-plans.md implementation or bugfix execution: references/test-driven-development.md debugging and root-cause analysis: references/systematic-debugging.md pre-merge or task-level review: references/requesting-code-review.md before saying work is done: references/verification-before-completion.md executing an approved plan with agent delegation: references/subagent-driven-development.md Do not bulk-read all references unless the task truly spans multiple workflow stages. Step 3: Reconcile with host rules before acting Before following any bundled superpowers guidance, reconcile it with: the user's explicit request, the current repository's AGENTS.md , the host runtime's tool and permission model, and any more specific installed skill that directly matches the task. If there is a conflict, follow the higher-priority local rule and use the bundled superpowers material as methodology, not authority. Step 4: Execute with honest boundaries Prefer the project's existing validation and review commands when they already exist. If the current repository has stricter branching, testing, review, or release rules, follow those. If the environment lacks multi-agent support, skip the subagent path and continue with the non-delegated workflow. If the task is trivial, keep the workflow compact rather than theatrical, but still choose an explicit workflow. Workflow Routing Guidance New feature, refactor, or behavior change Start with: references/brainstorming.md If the work becomes multi-step or cross-module, continue with: references/writing-plans.md During implementation, use: references/test-driven-development.md Before declaring success, use: references/verification-before-completion.md Bug, failing test, or unexpected behavior Start with: references/systematic-debugging.md Once root cause is understood and a fix is ready, use: references/test-driven-development.md Before closing the task, use: references/verification-before-completion.md Review or handoff checkpoint Use: references/requesting-code-review.md If the work was executed from a formal plan and the runtime supports delegation, optionally use: references/subagent-driven-development.md Safe Usage Rules Treat the bundled references as workflow assets, not as license to ignore local project rules. Prefer the smallest process that still makes the work explicit and honest. Do not claim TDD happened unless a failing test was observed first. Do not claim a bug is fixed unless the original failure mode was re-verified. Do not claim completion without fresh verification evidence. Do not require subagent usage when the current runtime, product tier, or user request does not support it. Troubleshooting Reference files missing Re-check the skill installation. Confirm skills/superpowers/ exists in the active workspace. Reinstall the marketplace skill if needed instead of inventing fallback paths. The current project already has stricter workflow rules Follow the stricter local rule. Use superpowers references only for the missing part of the workflow. The task needs the upstream standalone ecosystem Be explicit that the marketplace skill is a NextClaw adaptation, not the original multi-plugin install. Point the user to the upstream repository and install docs only if they want that exact standalone setup. The user wants speed over process Compress the workflow, but still keep the key gates: clarify intent for creative work, find root cause before fixing bugs, verify before claiming completion. Success Criteria This skill is working correctly when: the user gets an explicit workflow instead of ad-hoc thrashing, the smallest relevant superpowers reference is loaded for the task, local project rules remain the source of truth, debugging starts with root-cause investigation, implementation uses test-first discipline when appropriate, and completion claims are backed by fresh verification evidence. Attribution This skill adapts the upstream obra/superpowers project for the NextClaw marketplace. Upstream source mapping is documented in references/SOURCES.md .
このスキルを起動するキーワード。クリックでコピーできます。

このスキルにはトリガーワードがありません。

ダウンロードした .skill に含まれるフィールド。
フィールド 説明
formatフォーマット識別子(skill/v1)
skill_idスキル固有 ID
nameスキル名
versionバージョン
description説明
categoryカテゴリ(配列)
trigger_wordsトリガーワード
tagsタグ
sourceソース
source_urlソース URL(本ページ)
exported_atエクスポート日時(ダウンロード毎)
system_promptシステムプロンプト本文
model_configモデル設定:provider / model / temperature / max_tokens / top_p
examplesサンプル
install_guide各プラットフォームの導入説明(Coze / Dify / Claude / カスタム)
同じスキルを各プラットフォーム形式で出力できます。
.skill 標準形式。system_prompt と model_config を収録し、任意の Agent で利用可能 ダウンロード
.skillpro 拡張形式。scripts / tools / dependencies / hooks を含む ダウンロード
.json 純粋な JSON 出力。system_prompt とモデル設定のみ ダウンロード
Coze frontmatter 付き Markdown。Coze へのインポート用 ダウンロード
Dify Dify DSL。アプリ作成後にそのままインポート ダウンロード

每日精选 Skill 推荐,免费送到你邮箱

输入邮箱,每天接收一个精选 AI Agent 技能推荐。完全免费,持续更新。

验证码 --

提交后我们会发送一封确认邮件,点击邮件里的链接才会开始收信。

完全免费,取消任意时间。我们不会发送垃圾邮件。