本文へ移動
cccskills
無料GitHub で公開

brand-voice-analyzer

Analyzes a company's content to extract and codify their brand voice into a comprehensive style guide. Reads website copy, blog posts, emails, and social media to identify tone, vocabulary patterns, sentence structure, personality traits, and word preferences. Generates a brand-voice-guide.md and reviews new content against it.

インストール方法を見る

含まれるファイル(6)

  • SKILL.md3.8 KB
  • references/edge-cases.md2.7 KB
  • references/guide-template.md2.7 KB
  • references/metrics.md1.9 KB
  • references/review-mode.md1.7 KB
  • references/voice-dimensions.md3.8 KB

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。

Brand Voice Analyzer

Ingest a company's existing content, reverse-engineer the voice behind it, produce a definitive brand voice guide any writer can follow, and review new content against an established guide.

Contents

  • references/voice-dimensions.md -- the seven-axis tone spectrum, vocabulary, sentence, personality, audience, and formatting dimensions to extract
  • references/guide-template.md -- required sections, quality standards, and pre-delivery checklist for brand-voice-guide.md
  • references/review-mode.md -- process and output format for scoring content against a guide
  • references/metrics.md -- quantitative and qualitative analysis techniques and citation standards
  • references/edge-cases.md -- handling for thin corpora, multiple voices, regulated industries, rebrands, plus interaction patterns

Workflow: Create a New Voice Guide

  1. Ask which content sources are available (URLs, files, directories, pasted text).
  2. Ask about any known brand attributes, values, or existing guidelines.
  3. Collect the corpus. Pull URLs with WebFetch, locate files with Glob/Read, find public content with WebSearch. Gather website copy, 5-10 blog posts, email campaigns, social posts, sales collateral, and support content. Document each source and its word count.
  4. Confirm corpus size. Require at least 3,000 words across 2+ content types. If smaller, flag the limitation and mark affected findings as lower confidence.
  5. Run the full voice extraction across every dimension in references/voice-dimensions.md, citing source examples per the standards in references/metrics.md.
  6. Present a summary of findings for validation before generating the guide.
  7. Generate the complete brand-voice-guide.md following references/guide-template.md. Verify it against that checklist before delivery.
  8. Offer to review a sample piece of content against the new guide.

Workflow: Review Content

  1. Locate or ask for the brand-voice-guide.md.
  2. Read the content to be reviewed.
  3. Run the review process and produce the scored output defined in references/review-mode.md.
  4. Present findings with actionable, guide-referenced fixes.

Workflow: Update an Existing Guide

  1. Read the current brand-voice-guide.md.
  2. Ingest the new content sources and run extraction on the new material.
  3. Compare findings against the existing guide.
  4. Propose specific updates with rationale.
  5. Apply approved changes with Edit.

Output Standards

  • Primary output: brand-voice-guide.md. For multiple brands: brand-voice-guide-[company-name].md. Deliver review output inline unless a file is requested.
  • Use markdown headers (H1-H4), tables for comparative data, blockquotes for source quotes, code blocks only for literal reproduced text. Bold key terms on first use. No emojis.

Operating Principles

  • Never fabricate examples. Every quoted example must come from actual source content.
  • Never assert voice attributes without evidence. If the data does not support a conclusion, say so.
  • Present findings as observations, not prescriptions, until the user validates them.
  • Distinguish voice (consistent: who we are) from tone (shifts by context: how we adapt) in every guide.
  • When reviewing, be specific: cite the exact word, phrase, or structure and the guide section it touches.
  • Treat the guide as a living document and defer to the user's judgment on subjective calls.

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

Audit websites for accessibility issues and WCAG compliance. Use when checking accessibility, fixing a11y issues, or ensuring WCAG compliance.

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。

OneWave-AI/claude-skills3402026年10月2日 更新

Deploy a 2-layer parallel agent hierarchy for large, parallelizable work — big refactors, multi-file migrations, codebase-wide audits, bulk generation. A top-tier commander (Fable or Opus) orchestrates the swarms; the user picks a power level (Max Power / Heavy / Balanced / Economy) that sets the Opus/Sonnet/Haiku model mix per layer. Layer 1 is 3-50+ specialist agents, each with its own full context window; Layer 2 is 2+ sub-agents per member. Includes git safety, tiered sizing, a pre-deploy gate, phantom-completion checks, and multi-wave follow-up.

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。

OneWave-AI/claude-skills3402026年10月2日 更新

Use to test whether an AI agent, prompt, or skill actually works before or after it ships — build a small eval set, score runs, and catch regressions when the prompt or model changes. Trigger when the user asks whether a prompt is good, wants to compare models or versions, sees inconsistent agent output, or is about to put an agent in front of customers.

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。

OneWave-AI/claude-skills3402026年10月2日 更新

Deploys swarms of sub-agents for massive parallel data processing tasks. Unlike agent-army (which is for code changes), this is for DATA tasks -- processing 1000 documents, analyzing datasets, bulk content generation. Configurable swarm size, task distribution, result aggregation, progress tracking, and error recovery.

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。

OneWave-AI/claude-skills3402026年10月2日 更新

Designs and deploys custom agent teams for specific business workflows. Interactive discovery of business processes, then generates complete team configurations with specialized agent roles, tool access, communication protocols, and handoff rules.

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。

OneWave-AI/claude-skills3402026年10月2日 更新

Agent-to-Agent (A2A) communication protocol. Connect two or more Claude agents that pass messages, share context, delegate tasks, and collaborate. Implements structured handoffs, shared memory, and multi-agent conversations.

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。

OneWave-AI/claude-skills3402026年10月2日 更新

OneWave-AI のスキルをすべて見る

このスキルの問題を報告する