本文へ移動
cccskills
無料GitHub で公開

wtf

Make AI coding-agent messages to humans clear and easy to follow. Use for every user-facing update, explanation, review, question, and handoff in this repository, or when the user asks to simplify or rewrite an agent message.

インストール方法を見る

含まれるファイル(2)

  • SKILL.md2.5 KB
  • agents/openai.yaml215 B

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Human-facing writing

Apply this guide to all prose written for people. This includes chat messages, status updates, reviews, questions, handoffs, and pull-request summaries. Preserve exact commands, identifiers, schemas, and required templates.

Make the message easy to understand

  • Assume the reader is a mid-level or senior software engineer with working AWS knowledge. Explain project-specific context and unfamiliar concepts. Do not assume the reader remembers the full session.
  • Lead with the outcome or the information the reader needs most. Add background only when it helps the reader understand a decision, consequence, or next step.
  • Use plain, conventional sentence structure. Prefer an explicit subject, active verb, and clear object.
  • Split dense sentences. If a sentence requires backtracking, shorten it or divide it.
  • Avoid detached afterthoughts after commas, such as “..., extracted.” Write a complete clause or a separate sentence.

Keep the structure purposeful

  • Group related information. Use sections only when each section answers a distinct reader question.
  • Make relationships between sections and steps obvious. Avoid a wall of text and avoid a collection of unrelated headings.
  • When a message asks several questions, keep them together, preferably at the end, instead of scattering them through the text or across several messages. This does not apply to the harness's native question tool: use that tool whenever the user has to decide something, and let it carry the questions.
  • In a handoff, state what changed, which checks ran, and what remains uncertain. Mention implementation details only when they help the reader understand or maintain the result.

Remove distracting language

  • Remove filler, repetition, unnecessary jargon, and needless qualifiers.
  • Avoid sycophancy and stock chatbot phrases such as “Certainly” or “Found the smoking gun.”
  • Avoid rhetorical, literary, promotional, and marketing language. State the fact directly.

When invoked manually without a new task, rewrite the previous agent message using this guide. If there is no previous message to rewrite, confirm the change in one short sentence. Do not recite the guide unless the user asks for it.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Run, diagnose, or manually test the private Stacktape Console locally, especially when local API/UI behavior must use the shared dev RDS data, Cognito users, AWS services, OAuth callbacks, or provider webhooks. Use for Console dev-mode setup, deciding between UI-only and full mode, shared-dev migrations, and GitHub/GitLab/Bitbucket end-to-end testing.

日本語の概要は準備中です。原文の説明を表示しています。

stacktape/stacktape812026年10月10日 更新

load

無料

Load context from an earlier Claude Code or Codex session, selected by conversation name or recency. Explicit invocation only: `/load` in Claude Code, `$load` in Codex. Never trigger this from ordinary wording such as "load" or "continue".

日本語の概要は準備中です。原文の説明を表示しています。

stacktape/stacktape812026年10月10日 更新

Redesign Stacktape's marketing website or Console through visual exploration, independent design critique, and browser verification. Use for substantial page, flow, or visual-system design work, including agent-led redesigns.

日本語の概要は準備中です。原文の説明を表示しています。

stacktape/stacktape812026年10月10日 更新

wtf

無料

Make AI coding-agent messages to humans clear and easy to follow. Use for every user-facing message in this repository, or when the user asks to simplify or rewrite an agent message.

日本語の概要は準備中です。原文の説明を表示しています。

stacktape/stacktape812026年10月10日 更新

stacktape のスキルをすべて見る

このスキルの問題を報告する