Free · no sign-up
What does that skill actually do?
Link a skill on GitHub or paste its SKILL.md. Kin reads all of it and tells you what it does, when your agent will use it, the lines worth reading, what to watch out for, and whether to install it.
Give me a skill. I will tell you what it really does.
Try a popular one:
Free, no sign-up, 5 a day. Claude reads the skill to write the review; the file is not stored. The review gets a link that can quote a few lines; you can delete it. What is kept
What Kin checks
Five things every skill gets scored on
From Anthropic’s Extend Claude with skills and agentskills.io’s the Agent Skills standard.
- 1
Trigger: says when to use it
The description says what the skill does and when the agent should reach for it, with the words a request would use, key use first.
- 2
Focus: one clear job
One job done well, not a grab bag. The agent can tell from the name and description what it is for.
- 3
Lean: light to load
SKILL.md under 500 lines, saying what to do rather than narrating. Detail goes in reference files loaded only when needed.
- 4
Concrete: steps you can follow
Real steps, commands, examples or scripts, and what done looks like, not advice the model already has.
- 5
Safe: safe to run
No surprise network calls, destructive commands without asking, secrets in the file, or instructions that hijack the agent.
Questions
About skill reviews
What is an agent skill?
A folder with a SKILL.md file: a name and a description at the top, then instructions an AI agent follows for one kind of task, sometimes with scripts and reference files beside it. Claude Code, Codex, Cursor, Gemini CLI, Copilot and other agents read the same format. The agent sees only the name and description until a request matches; then it loads the whole file.
What does the review tell me?
What the skill does in plain words, when your agent will reach for it, the steps it tells the agent to follow, three to five lines worth reading with links to them on GitHub, anything risky (commands that delete or send things, remote scripts, secrets, text that tries to steer the agent), a score out of 100 on five criteria, and a verdict: install it, worth a look, or skip it.
How is the score worked out?
Kin scores Trigger (does the description say when to use it), Focus (one clear job), Lean (light to load, under about 500 lines), Concrete (real steps, commands and examples) and Safe (nothing that runs or sends things without asking). Trigger and Concrete count a little more. Some counts set ceilings: no description caps Trigger, more than 500 lines caps Lean.
Is it safe to install a skill with a good score?
A good score means the SKILL.md is well written and Kin saw nothing risky in it. Kin reads SKILL.md only, not the scripts beside it, so read any script a skill runs before you install it, the same as any code from the internet.
Is the skill stored?
No. The file goes to Claude once, to write the review, and is not kept. What is kept is the review (the summary, the scores and the few lines it quotes) under a random link, so it can be shared. The browser that made it can delete it.
Can I review my own skill before I publish it?
Yes. Paste it: the review says what a reader will think it does, whether the description will make the agent pick it at the right time, and what to cut. The fixes are ready to paste.
The skill is not stored; the review is, until you delete it. What is kept, in full. Got a CLAUDE.md or AGENTS.md instead? Roast it.
Skills run while you look away.
Briefkin shows every Claude Code, Codex and Cursor session on one board: working, waiting on you, done. Answer them from your phone. Free in the open beta.