# AI use cases for Hermes Jev Skills (kerpopule/hermes-jev-skills)

Repo: [kerpopule/hermes-jev-skills](https://github.com/kerpopule/hermes-jev-skills) | Use cases: 4 | Page: https://dedupedai.com/use-cases/kerpopule-hermes-jev-skills/

A set of Jev-powered decision tools (model routing, memory filtering, skill selection, web screening, mailbox sorting, and more) that plug into Hermes agents and also install for Claude Code and Codex.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Cut per-turn model spend in a Hermes agent: https://dedupedai.com/use-cases/kerpopule-hermes-jev-skills/cut-hermes-model-spend-per-turn/
- Add cheaper lane subagents to Claude Code: https://dedupedai.com/use-cases/kerpopule-hermes-jev-skills/add-cheap-lane-subagents-to-claude-code/
- Screen web search results for hidden instructions: https://dedupedai.com/use-cases/kerpopule-hermes-jev-skills/screen-web-results-for-hidden-instructions/
- Sort a mailbox into lanes with Jev: https://dedupedai.com/use-cases/kerpopule-hermes-jev-skills/sort-mailbox-into-lanes/

## How to cut per-turn model spend in a Hermes agent with Hermes Jev Skills

Repo: [kerpopule/hermes-jev-skills](https://github.com/kerpopule/hermes-jev-skills) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Let Jev pick a cheaper model per turn, shadow mode first, then switch on.

**Who it is for:** A developer running a Hermes agent on a frontier model who wants lower token costs without losing answer quality.

**The problem:** Agents spend frontier-model tokens on small routing decisions such as which model should answer this turn.

**The outcome:** Each turn is routed to a model Jev judges good enough, and a shadow log shows what it would have done before anything switches.

### Steps

1. Run `python3 install.py` from the cloned repo folder.
2. Run `jev setup-key`, paste your key into the page it opens, then run `jev doctor`.
3. Run `jev models suggest --write` to draft routing pools from price bands.
4. In Hermes, run `/jev routing shadow` and review the log for a day.
5. When the log looks right, run `/jev routing on`.

### Prompt to paste into your agent

```text
Install the hermes-jev-skills repo from kerpopule on this machine, set up the key, and run `jev models suggest --write`. Enable `/jev routing shadow` for the <profile> profile and report the decisions logged so far, with tier, model and confidence for each. Verify that no model switch happened while in shadow mode.
```

### Caveat

A plugin can swap the model but not the provider connection, and the routing commands are for Hermes only.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/kerpopule-hermes-jev-skills/cut-hermes-model-spend-per-turn/

## How to add cheaper lane subagents to Claude Code with Hermes Jev Skills

Repo: [kerpopule/hermes-jev-skills](https://github.com/kerpopule/hermes-jev-skills) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Install four Jev lane subagents, each with its own model and effort, into Claude Code.

**Who it is for:** A Claude Code user who wants routine subtasks handled by smaller, cheaper lanes instead of the main model.

**The problem:** Claude Code runs every subtask on the same model, even when the work is small.

**The outcome:** Claude Code gains four lane subagents (jev-lane-small through jev-lane-escalate) and a delimited lanes block in its CLAUDE.md file.

### Steps

1. Run `python3 install.py --check` to see what the installer would change.
2. Run `python3 install.py` to install for every agent it finds, including Claude Code.
3. If you want to keep your CLAUDE.md file unchanged, run `python3 install.py --no-claude-md` instead. (inferred, check the README)
4. Run `jev doctor` to confirm the install.
5. Start a Claude Code session and check that the four jev-lane subagents are listed.

### Prompt to paste into your agent

```text
Use the hermes-jev-skills repo from kerpopule to install the four Claude Code lane subagents on this machine. Run `python3 install.py --check` first and report what it would change. Then run the install and confirm that jev-lane-small, jev-lane-escalate and the two between them exist in ~/.claude/agents.
```

### Caveat

The installer edits ~/.claude/CLAUDE.md, but the README says it backs the file up first and that --no-claude-md leaves it alone.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/kerpopule-hermes-jev-skills/add-cheap-lane-subagents-to-claude-code/

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

Repo: [kerpopule/hermes-jev-skills](https://github.com/kerpopule/hermes-jev-skills) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

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

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

**The problem:** A fetched web page can carry instructions aimed at the agent, and the agent may read and follow them.

**The outcome:** Results are logged in shadow mode first, then the flagged parts are withheld with a notice in their place.

### 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 to paste into your agent

```text
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.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/kerpopule-hermes-jev-skills/screen-web-results-for-hidden-instructions/

## How to sort a mailbox into lanes with Jev with Hermes Jev Skills

Repo: [kerpopule/hermes-jev-skills](https://github.com/kerpopule/hermes-jev-skills) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Sort an inbox export into needs-reply, updates, promotional, sales and spam lanes.

**Who it is for:** A person with a backlog of email who wants a quick first sort with per-message cost known in advance.

**The problem:** A long inbox takes time to scan, and most messages do not need a person's attention.

**The outcome:** A summary of counts per lane, the messages Jev was unsure about, and what the run cost.

### Steps

1. Run `jev setup-key`, paste your key into the page it opens, then run `jev doctor`.
2. Export your messages to inbox.json; the README does not give the file format, so check the expected structure with `jev mail --help` before exporting. (inferred, check the README)
3. Run `jev mail --file inbox.json` to sort the mailbox into lanes.
4. Run `jev mail --file inbox.json --summary` for counts per lane and the cost.

### Prompt to paste into your agent

```text
Using the hermes-jev-skills repo from kerpopule, sort the file <inbox.json> with `jev mail --file <inbox.json>` and then `--summary`. Report the count per lane, the messages Jev was unsure about and the total cost. Verify the lane counts add up to the number of messages in the file.
```

### Caveat

The README does not say how mail content is redacted before it goes to the Jev cloud API, so test with non-sensitive messages first.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/kerpopule-hermes-jev-skills/sort-mailbox-into-lanes/