JevMade Sign in
← Back to experiments

Agent tooling

hermes-jev-approvals

An experiment in letting Jev review commands that Hermes would otherwise ask you to approve.

Source screenshot of hermes-jev-approvals
SOURCE SCREENSHOTFull screenshot ↗

What it does

It changes the approval step, not the agent's other decisions.

How you can use it

Start by writing a simple guide that separates routine chores from risky tasks. Harmless jobs might include clearing temporary files or stopping browser windows. Risky tasks might involve changing live business data or shared company systems. Your developer can then install the tool and supply an access key that connects your setup to an evaluation service.

The tool reviews each proposed command by asking several safety questions at once, such as whether it reveals private passwords or sends data elsewhere. Built-in rules then decide whether to run it automatically, block it, or pause for your confirmation. Automated checks can make mistakes, and the service inspects written text rather than testing it in a separate safety pen.

Maker-reported (not independently measured by JevMade): 9.8x faster, 4.2x fewer interruptions, same commands, same code path · Six questions cost the same as one: 345 ms average across the 20-case suite, versus 439 ms

Primitives
choice, score, noul
Platform
Python
Added
Project created
GitHub stars
4 (snapshot, not live)

Keep this for later

Sign in to bookmark experiments, guides and videos, and keep notes only you can see.

Continue to sign in

We’ll bring you back to this listing.