jev
Skill128 starsChanged 29 days agoMITView on GitHub
Sets up, checks or changes the optional Jev judge, the TypeSafe decision model that ship-check asks about review findings (is this real harm, is a small one worth fixing now, what proof does a small fix need). Shows whether it is on for each repo, which key it uses, switches it on or off, and sends one test request. Runs only when the user types it.
- Reads credentials
32 other projects use a version of this.
Adopted by
Most recently taken up by
- fralapo/awesome-agent-skillstheir own version, first seen Nov 7, 2026
- Lu1sDV/skillsmdtheir own version, first seen Nov 4, 2026
- josecabralf/skillstheir own version, first seen Nov 1, 2026
- dagonet/claude-code-toolkittheir own version, first seen Nov 1, 2026
- connorch/skillstheir own version, first seen Oct 28, 2026
- GastonGelhorn/jevmatetheir own version, first seen Oct 25, 2026
Other versions32 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 32
- wuyoscar/jev-skill★ 665same name
- FrancoisChastel/jev-code★ 44same name
- BorisLeMeec/jev★ 38same name
- flc1125/skills★ 38same name
- pragma-sh/pragma★ 25same name
- ciberjohn/Hermes-Skills★ 17same name
- ratacat/ratacats-skills★ 9same name
- Vincent-the-gamer/skills★ 9same name
- AlexPEClub/Jev-Model-Router-Claude-Code★ 8same name
- rashedInt32/jev-mcp★ 8same name
22 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- fix-the-classSkill
- habit-wordsSkill
- hulkSkill
- premortemSkill
- quickSkill
- reviewSkill
- setup-first-passSkill
- sharpenSkill
- ship-checkSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.
0
Common
Most of the 28 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 50
Do they have a Jev API key?: If not, say where one comes from (TypeSafe's console, https://console.typesafe.ai/keys) and stop.
- Very specific: line 97
Before writing a file path, check git ignores it: git -C <its repo> check-ignore -q <file>.
- Very specific: line 102
Run jev status <repo> again for each group, then once per key:
“Common” is measured: kept by at least 6 in 10 of the 28 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
jev
Jev is TypeSafe's decision model: it picks one option from a list and says how sure it is.
With the Claude Code plugin, ship-check can ask it three things about a review's findings:
- harm: does this finding do real harm (money, data, a side effect done twice or sent wrong, security, legal, a crash, stuck work, or the change not doing its job)? Jev can make a finding real harm. It can never clear one the review named, and an answer below its confidence floor, or any failure, leaves the call to the rules.
- priority: for the list of smaller findings, which to recommend fixing now.
- proof: for a small fix the user picked, the least proof it needs (the repo's checks only, a unit test, a test against the real database or service, or a browser test). A real-harm fix always gets the full proof.
Without the judge, the rules make each call as written. Nothing here is needed to use first-pass.
What leaves the machine
For each finding asked about: its scenario, its worst case, who meets it and, for proof,
the fix in one line, after secrets, emails and phone numbers are blanked. They go to
TypeSafe's API (https://api.typesafe.ai/v1/systemone) with the key. TypeSafe charges per
word sent; check the current price on its models page (https://docs.typesafe.ai/models)
before telling the user a number. Its data terms: https://docs.typesafe.ai/legal.
1. Where it stands
For the current folder, and for one repo of each group the user keeps apart (two companies with separate accounts):
node "${CLAUDE_SKILL_DIR}/../../scripts/cli.mjs" jev status <repo>It prints on or off, why, where the key comes from (a variable name and a file, never the key) and the config file's path.
2. What the user wants
From what they typed (nothing: show the status and ask; "off" or "on": do that), or ask in one message:
- Do they have a Jev API key? If not, say where one comes from (TypeSafe's console, https://console.typesafe.ai/keys) and stop.
- Where is it set? The name of the variable that holds it, and whether it is in the environment or in an env file (the file's path; it must be one git ignores). Never ask for the key itself; if one is pasted, don't repeat it, and say it should be rotated.
- For which repos? One key for everything, or a key per group of repos when they belong to separate accounts (then every group is named, and a repo no entry names has the judge off).
3. Write the config
The config file is the path jev status printed (first-pass/jev.json in the Claude Code
config folder, ~/.claude unless CLAUDE_CONFIG_DIR says otherwise). Read it first if it
exists and change only what the user asked:
One key for everything:
{ "keys": [{ "env": "TYPESAFE_API_KEY" }] }A key per account:
{
"keys": [
{ "env": "COMPANY_A_TYPESAFE_API_KEY", "file": "/abs/path/company-a/.env.local", "repos": ["/abs/path/company-a", "/abs/path/company-a-site"] },
{ "env": "COMPANY_B_TYPESAFE_API_KEY", "repos": ["/abs/path/company-b"] }
]
}env: the variable's name. It is read from the environment first, then fromfile.file: optional, an env file holdingNAME=value.repos: absolute paths. The deepest one that holds the folder wins, and a checkout of a named repo elsewhere (a worktree, a clean checkout in the temp folder) counts as that repo, even inside a folder named for another key. An entry withoutreposcovers every folder, and is allowed only as the only entry: with separate accounts, a folder no entry names has the judge off, never another account's key.- With separate accounts, name each repo's own folder. Never name a folder that also holds another account's repos (the main folder above them all, say): a named folder covers every folder inside it, new repos included, so a repo cloned there later would get that key.
"off": trueswitches the judge off everywhere and keeps the rest."model"pins a version (defaultjev-latest); pin one when you have tuned against it."url": only to send somewhere other than TypeSafe's endpoint (https, or this machine for a test server);jev statusthen shows it. Leave it out otherwise, and remove one the user did not ask for.- Before writing a
filepath, check git ignores it:git -C <its repo> check-ignore -q <file>. If it is not ignored, say so and don't use it.
4. Check it works
Run jev status <repo> again for each group, then once per key:
node "${CLAUDE_SKILL_DIR}/../../scripts/cli.mjs" jev test <repo>It sends one made-up finding (no code, no customer data) and prints Jev's pick and how sure it was. A refused key or no answer prints why.
5. Report
Jev judge: <on for <repos> (key from <NAME>, in the environment | in <file>) | off, <why>>, one line per key
Verified: jev test → <Jev's pick and confidence, or the error>, one line per key
Config: <the file, and what changed>Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

