Skills
GitHub repo ↗Claude Code skills and agents I wrote for my own work, open-sourced under MIT.
I use these every day for design handoffs, git workflows, career docs, and more. Before publishing, I stripped out anything sensitive or client-specific, so what's here is safe to install and adapt for your own projects.
/plugin marketplace add cosmefae/claude-skillsAnd the ones I keep
Alongside these, I keep a set of private skills for myself. They're close to a clone of how I like to work day to day: the routines, the shortcuts, the judgment calls I've refined over years. They raise the quality of my output more than any single tool I've used. As I wrote in the README for this repo, I never accepted that designers don't build.
screen-contract
designTurns a Figma frame, screenshot, or brief into a structural JSON contract before any code generation, preventing visual drift and invented tokens or components.
write-handoff
docsWrites a concise HANDOFF.md for the current task: state, file refs, verify commands, next steps.
commit-guided
gitGuided git workflow — validate, commit with Conventional Commits, open a PR.
sync-references
docsAudits and syncs documentation references (README, llms.txt, AGENTS.md, SKILL.md, CHANGELOG.md), broken links, orphan files, skill-index drift.
figma-shadcn-hygiene
designFigma-only structural hygiene pass: naming, taxonomy, auto layout, token gaps. Produces a Hygiene Report for handoff.
figma-shadcn-visual-match
designFigma to shadcn to code with a visual fidelity gate: implements, screenshots both sides, classifies diffs, iterates to approval.
career-boost
careerResume/LinkedIn diagnosis and rewrite for ATS and recruiters, with fit scoring against a target job.
crisis-investing
financeBuilds and scores investment strategies for global crisis scenarios: wars, pandemics, recessions, tail-risk events.
standardize-md-skills
docsAudits and standardizes SKILL.md/project MD files against Anthropic conventions, always scoped to an explicit target directory.
ai-daily-brief
newsGenerates a daily brief of top AI-world news, deduped across a curated, user-editable source list.
security-auditor
securityRead-only security audit of a Claude Code environment: secrets, permissions, config hygiene.