Skip to content
Kiragu MainaSystems workspace
Systems onlineSend a brief
All tools

Context-Aware Sales Assistant

Turn a moment of uncertainty into a relevant question about fit, setup, or limits, with advice based on your actual product list.

Shop owners who want to help interested visitors work out whether a product suits their project.

Visual examples. Plain language. Go at your own pace.

Context-Aware Sales Assistant

Package preview. Illustrations use sample data.

Helpful chatExample
Product pageWhat it does · What you need
How can I help?

Same opening, whatever the page

The challenge“How can I help?” on every page
Illustration only. These are sample screens and values.

Slide 1 / 6

Interested, but stuck on one question

Imagine a visitor likes your product, then pauses at the setup requirements. “How can I help?” gives them another blank box when they are unsure where to start.

  • They may need to know whether their website can use it.
  • Another purchase prompt does not answer that question.
1 of 6

What this means for your project

Give your website a chat helper that understands the page someone is reading.

The question they almost did not ask.

A visitor has read your product page twice. They are interested, but unsure whether it will work with their site. A well-placed question, “Will this work for my project?”, gives them a way forward. The assistant uses the page they are reading to help them ask the question that matters.

How would I use this with AI?

A skill is a set of instructions and examples for your AI coding assistant. You bring your idea and your project. The skill gives the assistant a process to follow.

  1. Tell your AI what you want to build and share the package instructions.
  2. Work through the setup together. You may need accounts or tools listed below.
  3. Review the changes and test them before putting them on your live site.

This is a package you use while building, not a finished website or a service that runs it for you.

Not sure where to start? Ask about a project like yours.

Explore the example

Made-up example
  1. Read the page
  2. Suggest a question
  3. Answer using product facts

Starting point

A generic opening question

What changes

A question about the page you are reading

Try an article, the shop list, or a product preview to see how suggested questions change. The answers are prewritten examples.

Compare questions on a sample article, catalog, and product page. Suggestions use written rules and invented data.

See how it works

Step 1 of 6

Interested, but stuck on one question

Imagine a visitor likes your product, then pauses at the setup requirements. “How can I help?” gives them another blank box when they are unsure where to start.

Starting situationA reader opens an article about payments
Where it breaksA generic opening question gives no help with the article’s topic
Made-up example. Try the interactive preview to see what changes.
  • They may need to know whether their website can use it.
  • Another purchase prompt does not answer that question.
About this example: Inspect the sample

Step 2 of 6

Give the conversation a useful opening

Beside the setup details, offer a question such as “What would I need to install?” The visitor gets a way to explore the obstacle in front of them.

BeforeThe same question on every page
AfterA suggested question about the current page and product
Made-up example. Try the interactive preview to see what changes.
  • Suggest questions about this page and this product.
  • Help them judge fit, including when a free option is more suitable.
About this example: Inspect the sample

Step 3 of 6

Use the page they are already reading

The rules consider the visitor’s request, current page, and limited recent activity. Product advice draws on your catalog so the answer reflects what you actually offer.

  1. Read the current page
  2. Suggest a relevant question
  3. Check product facts
  4. Answer and suggest a suitable product
Steps shown in the example.
  • Read price and availability from the catalog.
  • Skip questions already chosen, pause between offers, and respect opt-out.
  • Page text cannot grant the AI extra tool permissions.
About this example: Inspect the sample

Step 4 of 6

Walk in your visitor’s shoes

Move from a sample article to the catalog, then to a product page. See how questions change as the visitor explores what to build and which product might fit.

What to tryTry a different sample page, then dismiss suggestions
Expected resultQuestions change with the page. Dismissing them stops further suggestions in this example
Made-up example. Try the interactive preview to see what changes.
  • Compare each suggested question with the page it belongs to.
  • Try a product that is not yet available and see how the advice handles it.
Try the example
About this example: Inspect the sample

Step 5 of 6

Build a helper with rules you can read

Give your coding AI the question rules, catalog examples, and sample visitor situations. Use the checks to examine advice before putting it in front of customers.

SKILL.md
Instructions for choosing questions and recommending suitable products
scripts/
Question rules and limits for an added AI provider
examples/
Sample products and visitor situations
tests/
Checks for factual advice, turning suggestions off, and failures
What the package contains. Check the release details for the exact file list.
  • Examples for failures, slow replies, and unsuitable recommendations
  • A connection point for an AI provider you choose to add
  • Consistent question labels and ways to measure which suggestions people use
About this example: Inspect the sample

Step 6 of 6

A helpful shop conversation needs your product knowledge

This fits a shop owner with a maintained catalog and visitors who need help choosing. You keep the facts current and decide which questions are useful.

What you need
  • An up-to-date product list with accurate prices, availability, contents, and limits
  • A website where the assistant can tell which page or product the visitor is viewing
  • Node.js 20 or later to run the included examples on your computer; check the package record for tested versions
  • Your own AI provider if you add generated answers; live provider connections are not included or verified
Limits to knowPage-based suggestions do not prove higher sales or mean the assistant learns by itself
Check what you need and what the product can do before choosing it.
  • Suggestions follow written rules; the system does not automatically learn from conversations.
  • Local examples need no paid AI service. Any live provider connection needs separate setup and has its own costs.
  • A useful question is not a guarantee of a sale.
About this example: Inspect the sample

What you receive

  • Rules that use the current page and a limited amount of recent browsing context
  • Product suggestions based on your catalog, with pauses between offers and a way to turn them off
  • A connection point for an AI provider, runnable examples, and failure recovery
  • Sample cases to check answers and connection points for measuring use
See the files in the package
SKILL.md
Instructions for choosing questions and recommending suitable products
scripts/
Question rules and limits for an added AI provider
examples/
Sample products and visitor situations
tests/
Checks for factual advice, turning suggestions off, and failures

What you need to get started

  • An up-to-date product list with accurate prices, availability, contents, and limits
  • A website where the assistant can tell which page or product the visitor is viewing
  • Node.js 20 or later to run the included examples on your computer; check the package record for tested versions
  • Your own AI provider if you add generated answers; live provider connections are not included or verified

What you can check here

Package version 1.0.0 / page version 3

  • Inspect the sample

    The interactive preview uses invented data to explain the steps. It does not show customer results or evidence of increased sales.

Will it work with your tools?

Codex and Node.jsExample setup only
Paid package version 1.0.0. Check the package compatibility record for the exact setups tested. Other tools, hosting services, and live provider connections need validation in your own setup.

Before you choose

Read the paid skill terms for permitted use, what your purchase includes, help, and refunds.

  • One purchase licenses one individual or organisation to use and modify this version in unlimited own and client projects, including commercial work. Do not resell the standalone package, sublicense it as a competing package, redistribute it publicly, or share it with unrelated purchasers. Later releases may be sold separately; no lifetime updates are promised.
  • Suggestions follow rules you can inspect. The assistant does not automatically learn from visitors.
  • Any added AI provider has separate usage costs. Product recommendations do not guarantee sales.
Does it train on my visitors?

No. This version uses written rules and does not automatically train an AI model on visitor conversations.

Can it recommend something free?

Yes. Relevance and fit take priority over suggesting a purchase.