Design and AI
From Prompts to Process: Building SaaS Landing Page Forge
I combined 24 specialist design capabilities into one AI skill that turns content, product screens, and brand context into editable V1 landing-page directions in Figma.

From Prompts to Process: Building SaaS Landing Page Forge
AI can generate a landing-page hero in a few minutes.
But a complete SaaS landing page requires more than a strong first viewport. It needs to understand the audience, product, content, conversion goal, visual identity, product screens, and responsive behaviour.
I wanted an AI workflow that could consider these parts together instead of generating disconnected sections.
That became SaaS Landing Page Forge, an open-source AI skill for creating responsive landing-page directions inside Figma.
The skill takes real content, product screenshots, brand guidelines or a brand.md file, a conversion goal, and a target Figma file. It guides the agent through content structure, creative direction, Figma production, responsive thinking, and design review.
The output is meant for brainstorming and V1 exploration. It gives the designer an editable starting point, not a final design that can be passed directly to development.
Human taste, judgment, review, and refinement remain necessary throughout the process.
The problem with one large prompt
I had already experimented with prompts for landing-page design.
The results could look impressive at first. But once I looked beyond the hero section, the problems became clearer.
The agent might:
Create a generic page structure
Rewrite approved content
Invent customer logos or metrics
Ignore the actual product interface
Design only for desktop
Use brand colours without understanding their purpose
Stop after the first acceptable output
The problem was not a lack of AI design capability. It was the lack of a dependable process.
A landing page involves strategy, content, conversion, visual design, product design, responsiveness, accessibility, and review. Making the prompt longer did not automatically make the result better.

Combining specialist skills into one workflow
Over time, I had collected specialist AI skills covering SaaS conversion, typography, UX, visual design, motion, product design, responsive QA, Figma operations, and creative direction.
Each skill was useful independently. Loading all of them together created another problem.
Instructions overlapped. Some contradicted each other. Others were written for frontend development or different design tools.
I did not want to copy everything into one enormous file.
Instead, I audited 24 selected sources, identified the capability each one contributed, removed repeated or incompatible instructions, and combined the useful principles into one standalone workflow.
The final skill brings together:
Strategy and conversion
UX and content hierarchy
Visual direction
Product design
Editable Figma production
Responsive behaviour
Purposeful motion
Accessibility and design review
The original 24 sources do not need to remain installed. SaaS Landing Page Forge contains its own workflow, references, and quality checks.
What the skill does
The workflow begins by checking what the user has already provided.
The ideal inputs are:
Landing-page copy
Product screenshots or existing Figma frames
Brand guidelines or a
brand.mdfileA target Figma file
The audience and product offer
One primary conversion goal
If something important is missing, the skill asks for it. If the gap can be handled safely, it records an assumption or creates a labelled placeholder.
It does not invent product capabilities, testimonials, metrics, pricing, customer logos, or compliance claims.
Before changing a Figma file, it also checks what the connected Figma MCP can actually do. The agent needs to confirm that it can read the file, inspect the existing system, create editable layers, and review the result afterward.
This matters because attempting an edit is not the same as successfully completing one.
From content to a V1 design direction
The skill first creates a simple page thesis:
What does the product do?
Who is it for?
Why should someone trust it?
What should the visitor do next?
It then organises the supplied content into a possible conversion path.
The page may include a hero, product proof, feature sections, workflows, use cases, testimonials, pricing, FAQs, and a final CTA. But it does not force every product into the same template.
Once the content structure is clear, the skill develops a visual direction based on the actual brand and product.
This is where a structured brand.md file becomes useful.
Connecting it with BrandPrint
BrandPrint reads the visual language of a live website and generates a portable brand.md file.
SaaS Landing Page Forge can use that file alongside the landing-page copy and product screens.
The workflow becomes:
Read an existing website with BrandPrint
Generate its
brand.mdAdd the content and product screenshots
Use SaaS Landing Page Forge to explore a landing-page direction in Figma
Review and refine the result as a designer
The brand file gives the agent clearer information about colour, typography, tone, spacing, and visual behaviour.
It is more useful than asking the agent to make something “clean,” “modern,” or “premium.”
Creating editable Figma concepts
The skill works through a connected Figma MCP and aims to create native, editable layers.
When possible, it reuses the existing design system. If no system is available, it can create a small local foundation using styles, variables, grids, spacing rules, auto layout, and reusable components.
It also considers desktop, tablet, and mobile behaviour.
Responsive design is not treated as scaling down the desktop frame. The skill considers what should stack, reorder, crop, simplify, or disappear at each size.
Motion can also be suggested when it helps explain the product or guide attention. However, motion annotations are specifications, not finished production animations.
Why this is a V1, not a finished design
This distinction is important.
SaaS Landing Page Forge can help a designer:
Explore page structures
Test creative directions
Turn rough inputs into a visual starting point
Compare responsive ideas
Build an editable V1 faster
Identify missing content or product evidence
But its output should not be treated as development-ready.
The agent does not fully understand every business decision, user behaviour, brand sensitivity, or technical constraint behind the project. It may create a sensible direction while still missing a detail that an experienced designer would notice immediately.
The designer still needs to:
Challenge the content hierarchy
Refine the visual composition
Check brand accuracy
Review accessibility
Validate responsive behaviour
Confirm product authenticity
Simplify unnecessary elements
Prepare the final design system and handoff
AI can reduce the time needed to reach a useful starting point. It does not remove the need for taste.
Why a skill instead of a mega-prompt?
A mega-prompt tries to explain everything in one large instruction.
A skill provides a reusable process with stages, references, priorities, and review points.
SaaS Landing Page Forge can check the available inputs, inspect the connected tools, create a direction, build a first version, and review its own output.
It also has a clear priority order:
The user’s instructions
Supplied brand and product evidence
The existing Figma system
Factual accuracy and accessibility
Project-specific requirements
General design advice
This helps resolve conflicts between different design approaches without treating every instruction as equally important.
Making the skill portable
The package uses the SKILL.md format, allowing the same core workflow to work across compatible environments such as Claude, Cursor, and Codex.
It includes:
The main workflow
Supporting design references
Compatibility metadata
Installation guidance
Example prompts
Security notes
A source audit
MIT licensing
The Figma MCP remains separate. The skill does not contain access to anyone’s Figma account, credentials, or product files.
Each user controls their own tools, permissions, content, and brand material.
What building it taught me
The project changed how I think about AI-assisted design.
More instructions do not always create better work. The agent needs a clear process and a way to resolve conflicting advice.
Brand context is also more useful than style adjectives. A structured brand file gives the agent actual rules instead of asking it to guess what “premium” means.
Most importantly, review must remain part of creation.
The first output is a draft. Human judgment is what turns that draft into a thoughtful design.
From a personal workflow to an open-source skill
SaaS Landing Page Forge began as a way to make my own landing-page workflow more structured.
It is now available as an open-source project on GitHub and has also been listed on Agensi, a marketplace for agent skills.
Making it public meant documenting the workflow, installation, compatibility, limitations, security model, example prompts, and source decisions.
The skill is not a replacement for a designer or a shortcut to a final website. It is a structured brainstorming partner that helps designers move from raw material to an editable V1 with more clarity.

More of my AI experiments
SaaS Landing Page Forge is one part of how I have been exploring AI as a designer.
I have also been experimenting with brand-context tools, interactive product experiences, website workflows, Figma automation, coding agents, and smaller ideas that connect design with working technology.
I have collected these projects and experiments in one place, including the process, tools, and thinking behind them.
Final closing
The future of AI-assisted design is not one magic prompt.
It is a clear process, useful context, capable tools, and human taste guiding the final result.
You can explore the complete project through:

