jev-awesome-skills
Skill2 starsChanged 55 days agoMITView on GitHub
Use when a coding agent must make a closed choice, noul (yes/no), or score judgment with TypeSafe Jev instead of guessing. Load it to proceed, ask, or stop; to route a skill; to guard a command; to verify a claim; to triage a record; to review a diff; or to pick the next legal action. Jev returns probabilities only. The host still writes the code and checks the result. Skip it for open-ended writing, or when one file or one command already answers the question.
- Reads credentials
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- llms-full.txtllms.txt
- llms.txtllms.txt
- jev-actSkill
- jev-gateSkill
- jev-guardSkill
- jev-reviewSkill
- jev-routeSkill
- jev-triageSkill
- jev-verifySkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 1 is distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 17
Every skill in this package runs @aitofy/jev-awesome-skills.
- Very specific: line 45
If the key is missing, the transport throws, or the body is not Jev answers, the command returns "source": "fallback" and no answer …
- Very specific: line 49
Read branch when the answers include pick, determined, and harm.
- Very specific: line 52
proceed: when harm is below 1.5, the probability of the chosen pick label is at or above 0.9, and determined.noul is at or above 0.9.
- Very specific: line 55
Use the probability of the chosen label, not confidence.
- Distinctive: line 57
If one file or one command already answers it, proceed without calling Jev and without asking.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Jev awesome skills
Jev answers a closed question. It does not edit the repo, invent a plan, or replace a test. You write the code and check the result.
Every skill in this package runs @aitofy/jev-awesome-skills. Do not write a second client.
Catalog
| Job | Skill | Use it when |
| Policy | jev-awesome-skills | You need the cutoff rules below |
| Gate | jev-gate | A closed choice may already be determined by the repo |
| Route | jev-route | More than one installed skill could handle the turn |
| Guard | jev-guard | A command might destroy data, leak a secret, or leave the task |
| Verify | jev-verify | You are about to report a claim or "done" |
| Triage | jev-triage | A record needs one label |
| Review | jev-review | A diff needs ship, fix, or a person |
| Act | jev-act | The next step must be one of the actions you already listed |
Ask
jev questions gate --option pnpm="pnpm-lock.yaml is present" --option ask_user="not determined"
jev --dry-run --file request.json
jev --file request.json
jev --fixture examples/gate.response.jsonDry-run and --fixture do not open a socket. A live call reads the API key from the environment variable TYPESAFE_API_KEY. Never print the key. Never commit it. Never ask the user to paste it into chat.
Fail open
If the key is missing, the transport throws, or the body is not Jev answers, the command returns "source": "fallback" and no answer probabilities. Fail open: do not invent a probability and call it Jev. Decide yourself, or ask the user.
Proceed, ask, or stop
Read branch when the answers include pick, determined, and harm.
- stop when
harm.scoreis at or above 1.5. That is the midpoint between "changes behavior but is testable" and "may destroy data, leak a secret, or go beyond the task". Stop wins. Ask the user before any command that can do that, even if the other probabilities are high. - proceed when harm is below 1.5, the probability of the chosen
picklabel is at or above 0.9, anddetermined.noulis at or above 0.9. The context already determines the answer. Do not ask the user. Tell them what you inferred in one line. - ask otherwise. The question is undetermined.
Use the probability of the chosen label, not confidence.
If one file or one command already answers it, proceed without calling Jev and without asking. If you already know the command can destroy data, leak a secret, or leave the task, stop for confirmation even before calling Jev.
A proceed branch is not a passing test. You still write the change and run the checks.
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.