Skip to content

04-examples.md

Latest commit

 

History

History
195 lines (135 loc) · 5.82 KB

04-examples.md

File metadata and controls

195 lines (135 loc) · 5.82 KB

4. Real Examples

Time: 5 min · Goal: See Claude do something useful for your actual work.

Each example below shows the flow and what your screen will look like. Pick one and try it today.

Marketing or content work: start here. Working on code? The same patterns apply – see the official docs for code-flavored examples.

Example 1 – Find themes in customer feedback

Use case: 200+ NPS comments, survey replies, or support tickets. You need three clean themes for tomorrow's leadership update.

sequenceDiagram
    participant You
    participant Folder
    participant Claude
    You->>Folder: drop feedback.csv
    You->>Claude: claude
    You->>Claude: "find top 3 themes with sample quotes"
    Claude->>Folder: read feedback.csv
    Folder-->>Claude: 412 rows
    Claude-->>You: ranked themes + verbatim quotes
Loading

On your screen:

Terminal

~/feedback $ claude

  > Read feedback.csv. Find the top 3 themes.
    For each: 2 verbatim quotes + suggested fix.

  Reading 412 rows... done.

  1. Onboarding friction (87 mentions)
     "Took 4 days to figure out the import flow."
     "Wish there was a setup wizard."
     → suggest: build a guided first-run.

  2. Pricing confusion (54)
  3. Missing export (38)
  > _

Example 2 – Turn one blog post into ten social posts

Use case: You wrote a 1,500-word article. You need LinkedIn, Twitter, and a newsletter blurb – all in your brand voice.

flowchart LR
    Blog[blog/launch.md] --> Claude
    CL[CLAUDE.md<br/>brand voice] --> Claude
    Claude --> LI[1 LinkedIn post]
    Claude --> TW[5 tweets]
    Claude --> NL[1 newsletter blurb]
Loading

On your screen:

Terminal

  > Read blog/launch.md. Use CLAUDE.md voice rules.
    Generate:
    – 1 LinkedIn post (under 200 words, 1 emoji)
    – 5 tweets (under 250 chars, no emojis)
    – 1 newsletter blurb (3 sentences, no CTA)

  Drafts ready. Save them under /social/ ?  [y/N]

Press y and you have a full week of social ready to schedule.

Example 3 – Draft headlines that already sound like you

Use case: Three headline variants for a feature launch, on-brand, under 10 words.

On your screen:

Terminal

  > Write 3 headline variants for our "focus mode"
    launch. Use CLAUDE.md voice rules.
    Sentence case. Under 10 words each.
    Try 3 angles: benefit, skeptical, story.

  1. Stop pretending notifications aren't the problem.
  2. Focus mode: one button, your afternoon back.
  3. We tried it for a week. You should too.
  > _

The point: Claude already knows your voice from CLAUDE.md, so you don't have to re-explain it every time.

Example 4 – Tone-check and tighten existing copy

Use case: A draft mentions "leverage" and "ecosystem". You want a clean edit, not a rewrite.

flowchart LR
    A[post-draft.md] --> Claude
    B[CLAUDE.md<br/>banned phrases] --> Claude
    Claude --> C[Flagged lines<br/>+ replacements]
    C --> D{Apply all?}
    D -->|y| E[Edited file]
    D -->|n| F[Keep originals]
Loading

On your screen:

Terminal

  > Read post-draft.md. Flag every banned phrase
    from CLAUDE.md. Propose a one-line replacement
    for each. Don't change voice or structure.

  - Line 12: "leverage our ecosystem"
             → "use what we built"
  - Line 19: "world-class team"
             → "team that's been there"
  - Line 27: "game-changer"
             → "real shift"

  Apply all? [y/N] _

Example 5 – Translate without losing your voice

Use case: Localize an email sequence to French and Spanish, brand voice preserved, no machine-translation feel.

On your screen:

Terminal

  > Translate emails/welcome.md into French
    (formal "vous") and Spanish (neutral).
    Keep tone playful. Don't translate brand name
    or product feature names.

  ✓ emails/welcome.fr.md
  ✓ emails/welcome.es.md
  ✓ emails/glossary.md  (preserved terms)

  Want me to translate the rest of /emails/ too?
  > _

Make it reusable: turn examples into skills

Each prompt above can become a one-line shortcut you (or your team) reuse forever. Save the prompt once as a skill, then call it by name from any project:

flowchart LR
    P["Found a prompt<br/>that works"] --> Save["Save once as a skill<br/>~/.claude/skills/&lt;name&gt;.md"]
    Save --> Reuse["Reuse anywhere:<br/>'use &lt;name&gt; skill'"]
    Reuse -.same prompt, every time.-> P
Loading

A skill is a tiny markdown file describing when to use it and what to do. The five examples above are obvious candidates: find-themes, repurpose-blog, headline-variants, tone-check, localize. Build them once, your whole team gets the same on-brand output every time.

Learn more:

What's next?

You've seen the loop: open a folder → run claude → ask in plain English. Once you're comfortable:

  • Skills – save reusable instructions ("turn any blog post into social") that Claude can invoke automatically.
  • Hooks – run shell commands on events (e.g. auto-tone-check before commit).
  • MCP servers – connect Claude to your CMS, analytics, or CRM.
  • Subagents – delegate isolated tasks (research, drafting, fact-checking) in parallel.

Official docs: docs.claude.com/en/docs/claude-code.