The only four things that matter

1. Say what you want to create

Bad: "Help me with content." Good: "Write me 3 TikTok scripts about productivity hacks for founders. 30 seconds each."

You don't need to explain how — Vamos knows how to write scripts. You just need to tell it the output.

2. Name the platform and format

Different platforms, different rules. Always include:

Platform: TikTok, Reels, YouTube Shorts, YouTube long-form, LinkedIn, X/Twitter, Instagram caption, email, blog, ad copy, etc.
Format: script, caption, thread, hook, full post, ad copy

"Write me a 45-second TikTok script about X" > "Write me some content about X."

3. Give it a volume target

"Write me 10 TikTok scripts."
"Give me 30 hooks."
"Create 5 ad variations."

Vamos handles batch requests beautifully — each script gets its own Canvas for individual editing. Ask for more than you need and throw out the ones that don't land.

4. Chain research + writing in one prompt

Don't split: "Research X," → wait → "Now write scripts."

Say it all upfront:

"Analyze @competitor's top 10 Reels from the past month. Find the hook patterns. Then write me 5 scripts using the winning patterns in my voice."

Vamos uses cheap models for the research phase and switches to Claude for the writing — saving credits.

Teach Vamos what you don't like

The fastest way to improve outputs is to correct them in the moment. If something's off, just say it:

"I hate this hook. Never write anything like it again."
"Too corporate. I'd never say 'leverage' or 'unlock' — save that to memory."
"The CTA here is wrong. I always end with 'reply to this post' not 'click the link'."
"Stop starting scripts with a question. Use a statement."

Vamos saves the rule to Memory and applies it going forward. The more you use the app and give feedback, the sharper outputs get — across every future chat.

Good prompt examples

Simple ask (best for most tasks)

"Write me 5 Instagram captions about launching a new product. 50–80 words each."

Short, specific, clear output target. Vamos handles the rest.

Research + create

"Search trending videos about finance tips on TikTok in the past 2 weeks. Analyze the top 10. Then write me 5 scripts using the winning patterns in my voice."

One prompt, three phases.

Targeted research

"Analyze @garyvee's last 20 Instagram posts. What hook frameworks does he use most? Write me 3 hooks in a similar style but for business coaching."

Specific source, specific output.

Batch with variety

"Create 30 pieces of content: 10 TikTok scripts, 10 Instagram captions, and 10 LinkedIn posts — all about [topic]."

Iterative refinement

After the first output:

"Make script #3 shorter — 20 seconds — and change the hook to be more contrarian."

Or highlight a section in Canvas and say "rewrite this line more punchy."

Bad prompt examples (and the fixes)

❌ Too vague

"Help me with content." → Fix: "Write me 5 TikTok scripts about my niche [niche], each 30 seconds."

❌ Over-engineered

"Act as a senior content strategist with 10 years of experience. Using the following framework: [5 bullet points]. Consider my brand voice characterized by [3 paragraphs]..." → Fix: "Write me 5 TikTok scripts about X." (Your Brand Voice already loads. Skip the roleplay.)

❌ Micromanaging research

"Search for videos. Then pause and let me pick which to analyze." → Fix: "Find the top 10 videos on X and analyze them."

❌ Asking for "content" with no format

"Write content about morning routines." → Fix: "Write a 45-second TikTok script about morning routines."

Pro tips

Start generic, tighten after. Ask for 10 variations, then say "Redo #3 with a more emotional hook."
Use Memory for durable preferences. If you always want 30-second scripts, tell Vamos once: "Save to memory — I always want TikTok scripts at 30 seconds."
Tell Vamos what to avoid. "Never use the word 'unlock.'" "No emojis in captions." "Never start with a question."
Chain steps, don't split them. One long prompt > 3 short prompts with context loss.
Start a new chat when topic changes. Long chats accumulate context that can drag later outputs in weird directions.