DEDUPED Hot AI repos, turned into use cases Get 5 a day

How to screen web search results for hidden instructions with Hermes Jev Skills

Withhold parts of web results that carry instructions aimed at the agent.

Who it is forA Hermes user whose agent runs web_search or web_extract and who worries that pages may contain prompt injections.

The problemA fetched web page can carry instructions aimed at the agent, and the agent may read and follow them.

The outcomeResults are logged in shadow mode first, then the flagged parts are withheld with a notice in their place.

What you need first

  • Python 3.9 or newer (no dependencies)
  • A TypeSafe key from console.typesafe.ai
  • Hermes, Claude Code or Codex installed on the machine (the installer finds them)

Steps

  1. Run python3 install.py and jev setup-key if you have not already.
  2. In Hermes, run /jev screen shadow and let the agent use web tools for a while.
  3. Review the log of what would have been withheld.
  4. Run /jev screen on to withhold the flagged parts.

Prompt for your agent

On this Hermes machine, enable `/jev screen shadow` for the <profile> profile using the hermes-jev-skills repo from kerpopule. Run a web_search and a web_extract, then report which chunks would have been withheld and why. Confirm that the local pattern screen still runs if Jev is unreachable.

Caveat. The README reports 70 of 79 planted attacks caught in one test set and calls this not portable recall, and browser pages are never screened by this feature.

Share: X LinkedIn Hacker News Email Copy link

Disclaimer

Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page