Prompt Templates
Customize the AI prompts Borker uses to generate content for each platform and content type.
Prompt templates control the task instruction given to Claude during content generation. Your brand voice and voice attributes are the standing instructions (injected as system prompt). The prompt template is the specific request.
Manage templates in Settings → Prompts.
What prompt templates do
Every generation call to Claude has this structure:
[System prompt: brand voice + voice attributes]
[Prompt template: content type format + generation instructions]
[Context: topic, angle, platform, recent posts]The prompt template fills the middle section. It tells Claude what format to produce, how long it should be, what structure to follow, and any special instructions for that content type.
The 5 built-in templates
Borker ships with default templates for all five content types:
| Content type | Default template covers |
|---|---|
x-post | Single tweet ≤280 chars, hook-forward, no hashtag spam |
x-thread | Numbered format (1/ ... \n\n2/ ...), 4–6 tweets, hook + body + CTA |
linkedin-post | Professional hook, 150–300 words, ends with question, no link in body |
farcaster-cast | ≤320 chars, web3-native audience, opinionated take |
paragraph-essay | 600–1200 words, titled, sectioned, intro + body + conclusion |
When to use defaults vs custom templates
Use defaults when:
- You're just getting started
- The output quality is acceptable and you don't have strong opinions about format
- You want to minimize configuration overhead
Create custom templates when:
- You want a specific structure the default doesn't produce (e.g., X threads that always end with a question and a resource link)
- Your content type has a house format (e.g., your LinkedIn posts always start with a bold claim)
- The default template isn't producing the right length or tone
Custom templates override the built-in for that content type. You can have one template per content type, and you can revert to the default at any time.
How to write an effective prompt template
A good template is specific about format, length, and structure, but leaves the content itself to the topic and brand voice.
Template anatomy:
Write a [content type] for [platform].
Format:
- [specific format requirements]
- [length requirement]
- [structure requirement]
Tone:
- [any tone specifics not covered by brand voice]
Do not:
- [specific things to avoid]Example: Custom X thread template
Write a Twitter/X thread on the provided topic.
Format:
- Number each tweet: "1/ ...\n\n2/ ..."
- 5 tweets total
- Tweet 1: contrarian hook that challenges an assumption
- Tweets 2–4: one specific insight or example each
- Tweet 5: one sentence summary + question for replies
Do not:
- Use vague openers like "Here's what I've learned..."
- End with "RT if you agree"
- Include hashtagsExample: Custom LinkedIn template
Write a LinkedIn post on the provided topic.
Format:
- Open with a single bold sentence (the most important point)
- 3–4 short paragraphs (2–3 sentences each)
- Close with a specific question to prompt comments
- No links in the post body
- 180–220 words totalWhere templates fit in the generation pipeline
graph LR
A[Brand Voice Config] --> D[System Prompt]
B[Voice Attributes] --> D
C[Prompt Template] --> E[User Prompt]
F[Topic + Angle] --> E
G[Platform Context] --> E
D --> H[Claude API]
E --> H
H --> I[Generated Content]The template is part of the user prompt (the specific task request). Brand voice is always in the system prompt. The AI uses both to generate content that's on-brand and on-format.
If content quality is consistently wrong in a specific way (always too long, wrong structure, wrong tone), the prompt template is usually the fastest fix. It's more targeted than adjusting brand voice.