JevMade field notes / Architecture and workflow guide

Adding safety checks to an AI assistant using a decision tool

This guide explains how to check an AI assistant's planned actions before they happen. It uses a tool to spot dangerous commands, hide private information, and handle errors without breaking the workflow.

Original by y0usafGuardrails

Bookmark Keep this in your collection.

Leave a noteWhat would you try with this?

Only you can see your notes.

Listen to this guide

JevMade’s plain-English explanation

0:00 /

AI narration

Independent ElevenLabs affiliate. JevMade may earn a commission.

Credits and license

“pi-jev (y0usaf)” by y0usaf. Read the original source.

This expanded guide is an AI-narrated adaptation of the source’s essential explanation, examples and caveats, not a word-for-word reading. The synthetic voice does not imitate the author or imply their endorsement.

License: MIT

MIT License

Copyright (c) 2026 y0usaf

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

Our summary

This guide explores an add-on for an AI assistant that writes code. It uses Jev, an AI tool that chooses from options rather than writing answers. People use this to review the assistant's planned actions and catch dangerous commands before anything runs.

The software asks specific questions about a planned task, like whether it deletes files or shares data. It shortens long text before sending it for review. After a task finishes, it checks the results to see if any API keys, which are private access codes, leaked.

This setup is useful for developers who want to monitor their AI assistant. If the safety check loses connection, the assistant continues working anyway. The author only tested these settings on a small number of examples, so the limits might not catch every problem.

Key takeaways

Cut down long text and remove sensitive details before sending data for an outside safety check.
Group different types of errors into categories to give the AI assistant clear instructions on fixing them.
Let the assistant keep working if the safety check fails, but limit how often errors are reported.

More guides →See the experiments →