Skills
Agents I built to design faster.
When you design by building, the tooling is part of the craft. These are skills I use every week — packaged so you can use them too. Each is a folder with a SKILL.md that any skill-compatible agent can load.
unzip name.skill -d ~/.claude/skills/or use “Save skill” when you open a .skill file in Claude./ask-clara
Ask an agent anything about me.
Drop it into Claude Code, Claude.ai, or any skill-compatible agent and ask how I work, what I've shipped, or whether I'd fit a role. It answers from a grounded profile and says when it doesn't know.
/iterate
Critique → three real options → converge. Repeat.
Takes a screenshot, wireframe, or a component in the repo and iterates in structured rounds against a stated bar. Refuses to pass off color swaps as exploration.
/synthetic-persona
A user built from real research, run against your product.
Builds a persona from interview transcripts or tickets — with their vocabulary, not yours — then tries to complete real tasks in a prototype or live product and reports where it fails. Flags every guess that's outside the research.
/critical-feedback
Hear the objection a day before the review.
Simulates an exec, legal, PMM, or a named stakeholder from their actual direction docs and past feedback. Ranked objections, what would change their mind, and the questions to have answers for.
/design-cleanup
Make a vibecoded prototype match the design system.
Finds the system, audits hardcoded colors, spacing, type, duplicate components and missing states (script included), reports before touching anything, then fixes in reviewable commits.
/design-brief
Turn a prototyping session into a decision document.
Pulls the design decisions, rationale, rejected options and trade-offs out of a repo, PR, or agent session and packages them with screenshots and a prototype link — ending with the exact approval you need.
How I use them together
A typical week: /iterate on a screen in the repo until it clears the bar. /synthetic-persona against the prototype to catch the obvious failures before I spend real users on them. /critical-feedback the night before a review so I've already heard the objection. /design-cleanup before the front end merges. /design-brief to get the decisions approved without a meeting.
None of them replace users, stakeholders, or engineers. They make the first draft of each conversation better.