jev-judgment
Agent skill that sends closed coding-agent judgments to Jev before asking the user, before risky commands and after failures.
Overview
jev-judgment is an Agent Skill, installable with the skills CLI and independent of any particular harness, that sends closed judgements to Jev and acts on the probabilities while the main coding model keeps writing code. It calls Jev at three points: before asking the user a closed question (if the conversation already determines the answer, proceed and say so), before a command that may destroy data, leak secrets or reach beyond the task (stop and confirm when any dimension crosses its threshold), and after a command fails (classify the failure and check output for secrets before retrying).
The README states Jev's confidence is evidence, not permission: deletion outside the request, breaking changes, new dependencies, paid or remote actions and anything the user asked to be consulted on always go to the user. The bundled standard-library Python script redacts credentials, truncates long strings and fails open on network errors or a missing key. Documentation is in English and Korean.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does the conversation already answer this closed question? | — |
| Score | Could this command destroy data, leak secrets or go beyond the task? | — |
| Choice | What caused this command failure? | — |
How it works
- Install the skill and set TYPESAFE_API_KEY
- Before closed questions, risky commands or after failures, call scripts/jev.py
- Compare answers with thresholds in SKILL.md
- Proceed, confirm with the user or retry; fail open on errors
Strengths
- Fewer unnecessary interruptions to the user
- Credential redaction before sending
- Harness-independent
Caveats
- Confidence is advisory only
- Thresholds are fixed in the skill and may need tuning
Getting started
`npx skills add HyunjunJeon/jev-judgment -g -a claude-code codex`, then set TYPESAFE_API_KEY.At a glance
- Type
- Model
- Status
- beta
- Language
- Python
- Runtime
- Agent skill
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Coding agents: agent skill that sends closed coding-agent judgments to Jev so verdicts stay typed, cheap, and comparable across runs.
yibie/awesome-jevAdds judgment checks for authorization, operation risk, and failure causes to coding Agents.
logicrw/awesome-jev-projectsAgent Skill: send closed coding-agent judgments to TypeSafe Jev.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Whether a closed question is already answered, how risky a command is, and why a command failed.
Can Jev approve risky actions?
No, certain actions always go to the user regardless of confidence.
What if I have no key?
The script fails open and the agent continues normally.
Related projects
- SystemOneHarness★ 193
Agent harness that compiles an environment's finite actions into typed questions for Jev and gates each step by confidence.
CLIbetaPython Accessibility-first macOS computer-use MCP server and CLI, with optional Jev guards on targets and inputs before irreversible actions.
PluginbetaPythonClaude Code skill that splits work into PBIs for Codex, Antigravity or Claude sub-agents, using Jev for model choice and triage.
ProjectprototypePythonClaude Code function hook that sends the agent's this-or-that questions to Jev and optionally ranks installed skills per prompt.
ProjectbetaTypeScriptSkill and CLI that compares a shortlist of agents, MCP servers or skills against a task and permission ceiling, abstaining on weak leads.
MCP serverprototypeTypeScriptCodex computer-use skill where Jev picks the next on-screen element, action, completion and risk from text candidates.
Agent skillprototypeJavaScript