anthropics/knowledge-work-plugins

discover-brand

This skill orchestrates autonomous discovery of brand materials across enterprise platforms (Notion, Confluence, Google Drive, Box, SharePoint, Figma, Gong, Granola, Slack).

소스 보기

사용 방법

  1. 설치 명령을 복사하고 자신의 에이전트 환경에 설치하기 전에 소스 패키지를 확인하세요.
  2. 필요한 도구, 인증 정보, 지원 환경은 원본 문서에서 확인하세요.
  3. 입력과 원하는 산출물을 설명하고 출처와 결과를 대조하세요.
npx skills add https://github.com/anthropics/knowledge-work-plugins --skill discover-brand

시작 전 확인

  • 설치에는 모델 구독, API 한도, 파일 접근 권한이 포함되지 않습니다.
  • 별점 수와 설치 수는 활동 지표이며 독립적인 품질·보안 검증이 아닙니다.
원본 Skill 문서

원본 저장소의 제목, 예시, 코드, 표, 링크, 이미지를 유지해 표시합니다.

Brand Discovery

Orchestrate autonomous discovery of brand materials across enterprise platforms. This skill coordinates the discover-brand agent to search connected platforms (Notion, Confluence, Google Drive, Box, Microsoft 365, Figma, Gong, Granola, Slack), triage sources, and produce a structured discovery report with open questions.

Discovery Workflow

0. Orient the User

Before starting, briefly explain what's about to happen so the user knows what to expect:

"Here's how brand discovery works:

  1. Search — I'll search your connected platforms (Notion, Google Drive, Slack, etc.) for brand-related materials: style guides, pitch decks, templates, transcripts, and more.
  2. Analyze — I'll categorize and rank what I find, pull the best sources, and produce a discovery report with what I found, any conflicts, and open questions.
  3. Generate guidelines — Once you've reviewed the report, I can generate a structured brand voice guideline document from the results.
  4. Save — Guidelines are saved to .claude/brand-voice-guidelines.md in your working folder once you approve them. Nothing is written until that step.

The search usually takes a few minutes depending on how many platforms are connected. Ready to get started?"

Wait for the user to confirm before proceeding. If they have questions about the process, answer them first.

1. Check Settings

Read .claude/brand-voice.local.md if it exists. Extract:

  • Company name
  • Which platforms are enabled (notion, confluence, google-drive, box, microsoft-365, figma, gong, granola, slack)
  • Search depth preference (standard or deep)
  • Max sources limit
  • Any known brand material locations listed under "Known Brand Materials"

If no settings file exists, proceed with all connected platforms and standard search depth.

2. Validate Platform Coverage

Before confirming scope, check which platforms are actually connected and classify them:

Document platforms (where brand guidelines, style guides, templates, and decks live):

  • Notion, Confluence, Google Drive, Box, Microsoft 365 (SharePoint/OneDrive)

Supplementary platforms (valuable for patterns, but not where brand docs are stored):

  • Slack, Gong, Granola, Figma

Apply these rules:

  1. If zero document platforms are connected: Stop. Tell the user: "You don't have any document storage platforms connected (Google Drive, SharePoint, Notion, Confluence, or Box). Brand guidelines and style guides almost always live on one of these. Please connect at least one before running discovery. Gong/Granola/Slack transcripts are valuable supplements but unlikely to contain formal brand documents."
  1. If no Google Drive AND no Microsoft 365 AND no Box: Warn (but proceed): "None of your primary file storage platforms (Google Drive, SharePoint, Box) are connected. Brand documents frequently live on these platforms. Discovery will proceed with [connected platforms], but results may have significant gaps. Consider connecting Google Drive or SharePoint."
  1. If only one platform total is connected: Warn (but proceed): "Only [platform] is connected. Discovery works best with 2+ platforms for cross-source validation. Results from a single platform will have lower confidence scores."

3. Confirm Scope with User

Before launching discovery, confirm:

  • Which platforms to search (default: all connected)
  • Whether to include conversation transcripts (Gong, Granola) or just documents
  • Any known locations to prioritize

Keep this brief — one question, not a questionnaire.

4. Delegate to Discover-Brand Agent

Launch the discover-brand agent via the Task tool. Provide:

  • Company name (from settings or user input)
  • Enabled platforms
  • Search depth
  • Any known URLs or locations to check first

The agent executes the 4-phase discovery algorithm autonomously:

  1. Broad Discovery — parallel searches across platforms
  2. Source Triage — categorize and rank sources
  3. Deep Fetch — retrieve and extract from top sources
  4. Discovery Report — structured output with open questions

5. Present Discovery Report

When the agent returns, present the report to the user with a summary:

  • Total sources found and analyzed
  • Key brand elements discovered
  • Any conflicts between sources
  • Open questions requiring team input

6. Offer Next Steps

After presenting the report, offer:

  1. Generate guidelines now — chain to /brand-voice:generate-guidelines using discovery report as input
  2. Resolve open questions first — work through high-priority questions before generating
  3. Save report — store the discovery report to Notion or as a local file
  4. Expand search — search additional platforms or deeper if coverage is low

Open Questions

Open questions arise when the discovery agent encounters ambiguity it cannot resolve:

  • Conflicting documents (e.g., 2023 style guide vs. 2024 brand update)
  • Missing critical sections (e.g., no social media guidelines found)
  • Inconsistent terminology across platforms

Every open question includes an agent recommendation. Present questions as "confirm or override" — not dead ends.

Integration with Other Skills

  • Guideline Generation: The discovery report is returned by the discover-brand agent via the Task tool. Pass it directly to the guideline-generation skill as structured input, replacing the need for users to manually gather sources.
  • Brand Voice Enforcement: Once guidelines are generated from discovery, enforcement uses them automatically.

Error Handling

  • If zero platforms are connected, inform the user which platforms the plugin supports and how to connect them.
  • If all searches return empty results, flag the discovery as "low coverage" and suggest the user provide documents manually or check platform connections.
  • If a platform is connected but returns permission errors, note the gap and continue with other platforms.

Reference Files

For detailed discovery patterns and algorithms, consult:

  • `references/search-strategies.md` — Platform-specific search queries, query patterns by platform, and tips for maximizing discovery coverage
  • `references/source-ranking.md` — Source category definitions, ranking algorithm weights, and triage decision criteria

관련 작업

같은 저장소의 Skills

더 많은 Skills

모든 Skills
anthropics
공식

data-visualization

Create effective data visualizations with Python (matplotlib, seaborn, plotly). Use when building charts, choosing the right chart type for a dataset, creating publication-quality figures, or applying design principles like accessibility and color theory.

설치 수
3
GitHub Stars
3.2만
업데이트
10월 22일
anthropics
공식

ux-copy

Write or review UX copy — microcopy, error messages, empty states, CTAs. Trigger with "write copy for", "what should this button say?", "review this error message", or when naming a CTA, wording a confirmation dialog, filling an empty state, or writing onboarding text.

설치 수
3
GitHub Stars
3.2만
업데이트
10월 22일
anthropics
공식

daily-briefing

Morning rundown - today's meetings with account context, opps closing soon with stale flags, waiting customer emails, and the day's top actions. Use when the user says "daily briefing", "daily brief", "morning briefing", "what's my day", "what's on my plate today", "prep my day", "start my day", "morning rundown", or on a schedule.

설치 수
2
GitHub Stars
3.2만
업데이트
10월 22일
anthropics
공식

incident-response

Run an incident response workflow — triage, communicate, and write postmortem. Trigger with "we have an incident", "production is down", an alert that needs severity assessment, a status update mid-incident, or when writing a blameless postmortem after resolution.

설치 수
2
GitHub Stars
3.2만
업데이트
10월 22일