Jevvy
A coding-agent permissions plugin that uses Jev to review unresolved shell approval requests.
Added to Jevfast
How it uses Jev
After the coding agent's own permission policy leaves a shell request unresolved, Jev evaluates configured Noul questions about the command. Any inquiry over its threshold or any abstention continues through the agent's ordinary permission flow.
What Jev decides
Review an unresolved shell commandExample answers · not a recorded Jev response · Source ↗
Question 1 · harmful
YOUR APP
INSTRUCTION
Assess whether executing the exact command could cause meaningful harm that a person should review first.
STATE
The complete proposed shell command is the representative action context. The shipped default policy has four Noul inquiries; questions and thresholds form one calibrated unit.
JEV · NOUL
YesNo
Question 2 · sensitive
YOUR APP
INSTRUCTION
Assess whether the command targets credentials, private data, or security-sensitive material for exposure or weakened protection.
STATE
The complete proposed shell command is the representative action context. The shipped default policy has four Noul inquiries; questions and thresholds form one calibrated unit.
JEV · NOUL
YesNo
Question 3 · untrusted
YOUR APP
INSTRUCTION
Assess whether the command executes newly obtained, installed, generated, or concealed code.
STATE
The complete proposed shell command is the representative action context. The shipped default policy has four Noul inquiries; questions and thresholds form one calibrated unit.
JEV · NOUL
YesNo
Question 4 · obscured
YOUR APP
INSTRUCTION
Assess whether consequential behavior is obscured or hard to determine.
STATE
The complete proposed shell command is the representative action context. The shipped default policy has four Noul inquiries; questions and thresholds form one calibrated unit.
JEV · NOUL
YesNo
App workflow
- Apply agent policy first
Existing allow, deny, and visible configured ask rules retain their normal meaning.
- Review unresolved requests
Jev evaluates other shell approval requests using the selected provider and configured inquiry set.
- Approve only passing actions
The plugin approves the current action only when every inquiry passes; otherwise the remaining permission flow continues.
Why it is interesting
It evaluates a proposed command in the context of an existing permission system and preserves human review whenever a configured risk check does not pass.