Free · runs on your computer · sends nothing
Every skill your agents load, on one page
Claude Code, Codex and Cursor each keep their own skills, plus a copy for every plugin version. One command shows them all: which agent sees what, what loads twice, and what you haven't used in a month.
Terminalcurl -fsSL https://briefkin.dev/skill-check | shMac or Linux with node 18+. No account. Reads only names and descriptions.
What it shows
Find the skills that cost you and the ones you forgot
Every skill, by agent
Your own skills, plugin skills, built-ins and the skills in the repositories you worked in, for Claude Code, Codex, Cursor, Gemini CLI, Copilot CLI, OpenCode, Windsurf and Amp.
Duplicates and versions
Skills an agent loads twice, like one plugin from two marketplaces, and copies that drifted apart because you edited only one.
What you actually used
Sessions in the last 30 days for each skill, counted from Claude Code transcripts and Codex session logs on your computer.
What the list costs
About how many tokens each agent spends on skill names and descriptions in every session, before you type a word.
Where each agent looks for skills, folder by folder: Where are Claude skills stored?
Questions
About the skill check
What does it send to Briefkin?
Nothing. The check runs on your computer and writes one HTML file, ~/.config/briefkin/skill-check.html, which loads nothing from the internet. Briefkin only counts how many times the script is downloaded, without your address or anything about your computer.
What does it read?
The skill folders of each agent, and from each SKILL.md only the name, description and version; plus the last 30 days of Claude Code transcripts (~/.claude/projects) and Codex session logs (~/.codex/sessions), looking only for which skills were loaded and called. It doesn't change, move or delete anything.
What do I need?
A Mac or Linux computer (or WSL on Windows) with node 18 or newer, which most coding agents already install. No account, no sign-up.
Why only Claude Code and Codex for use counts?
They are the agents that keep local logs saying which skill ran. For Cursor, Gemini CLI, Copilot CLI, OpenCode, Windsurf and Amp the page shows which skills they can load, but not how often they ran.
How do I try a skill or keep one handy?
Each card has the command that runs it, such as /unslop for Claude Code or $unslop for Codex, which you copy with a click. Star the ones you want to find again; stars stay in that browser.
Can I run it again later?
Yes, any time: it rewrites the page with what is there now. Add -s -- --days 60 after sh to count 60 days of use instead of 30.
Now see what your agents are doing
Briefkin puts every Claude Code, Codex and Cursor session on one board: working, waiting on you, done. Answer them from your phone, and see what each skill produced.
Free during the open beta. No card.