THE WORKBENCH / OPEN SKILLS

Good work.A place to start again.

The AI workflows I use for demos, page reviews and debugging. Shared with people who like making things, too.

Explore the skills ↗

03 public skills · MIT licensed

Made to be used, questioned and improved.

Find your starting point

Build an idea. Review a page. Understand a bug.

MARINA CURRENT ORIGINALPUBLIC

interactive-html-demo

Interactive HTML Demo

Turn an idea into a demo people can click, try and share. A single HTML file, with clear states and recovery paths.

INTERACTIVE WALKTHROUGHFrom an idea to a reviewable flow.Open / close the walkthrough

Local simulation · No AI request or real data is sent.

01Brief
02Flow
03Failure
04Recovery

Start with a decision

A team needs a way to review incoming ideas. The proposal must show how an idea becomes a decision.

WHAT CHANGESInput: a short brief → Output: a reviewable proposal

Who: a small project team
Main action: review an idea
Success: the decision is visible

VERSION: 1.0.0
LICENSE: MIT
VERIFICATION: 15 checks · 3 sizes

How to use ↗Original starter ↗GitHub ↗

REFERENCE-BASED REBUILDPUBLIC

web-design-qa

Web Design QA

After a page opens, check whether it reads well, works well and feels right on a phone.

FIELD NOTESAn idea worth trying

Record the context, choose a next step, and make room to review what happened.

Read the note ↗

After: a clear title, comfortable line spacing and a distinct reading link. The same content, easier to scan.

Illustrative example, not a client result. The reading link inside the preview is a label.

VERSION: 1.0.0
LICENSE: MIT
EVALS: 5

ARTICLE GITHUB

OPEN-SOURCE ADAPTATIONPUBLIC

diagnosing-bugs

Diagnosing Bugs

Find the evidence and a repeatable failure before deciding what to change. Adapted from Matt Pocock’s MIT-licensed workflow.

01
Reproduce
Record the trigger, expected result and actual behavior.
02
Locate
Use logs and a minimal example to test a hypothesis.
03
Retest
Repeat the original path and check adjacent behavior.

VERSION: 1.0.0
LICENSE: MIT
EVALS: 5

ARTICLE GITHUB

ON THE DESKWRITING IN PROGRESS

ai-native-product-playbook

AI-Native Product Playbook

I’m turning the transferable principles into an article. There is no public skill package or article link yet.

PUBLIC BOUNDARY

Share the method, protect the context.

Transferable principles will be shared as writing. Private operating rules and internal collaboration details stay private.

Publication format: article · In progress

Article → Skill → Evidence

The story explains the decision. The repository carries the reusable files. The release notes show what was actually checked.

READ THE FIRST STORY

The public standard

A Skill is listed as public only after three checks.

01
Source
Original, rebuilt and adapted work are labeled differently.
02
Boundary
Private data and internal operating rules stay out.
03
Proof
Examples, checks, limits and changes ship with the files.