Jev in production › Moderation and content filtering
JevPromptShield
Scores incoming text with Jev to flag prompt injection before an agent acts; a test on 21 published Gandalf jailbreak prompts blocked 6 of 21 (author).
6 of 21measured, as published by the source
- Use
- Moderation and content filtering
- Industry
- Security
- Form
- Open-source tool
- Stage
- In production
- Listed
- 2026-10-09
- Found via
- github
- Repository
- JamesANZ/JevPromptShield
- Stars
- 202
- Forks
- 46
- Last push
- 2026-10-08
- Language
- TypeScript
- License
- none stated
The README opens with
A prompt can tell a coding agent to drop its task and hand the secrets over. A pasted page can hide the same instruction in what looks like documentation. The agent can then propose a shell command that wipes a disk, force-pushes main, or pipes a downloaded script into bash.
JEV Shield stands outside the model and checks both moments. TypeSafe Jev scores the text. Ordinary code turns that score into a block, a question, or a pass. The agent does not get a vote.
Badge
For the project's own README, linking back here:
Also used for moderation and content filtering
- jev-edge from 37% to 5%
- slop-filter 19
- Sift About $0.00003 per post
- ElevenLabs 180ms
- kassad 0.987
- FakeCatch about a second
- Near Here 96 per cent
- ClubDoorman