Jev.aitools.fyi

Command Palette

Search for a command to run...

Featured
Product profile
Python381 starsUpdated 2026-09-21

Hermes Jev Skills

hermes-jev-skills (github.

Overview

hermes-jev-skills (github.com/kerpopule/hermes-jev-skills) packages nine Jev-powered skills as plain SKILL.md files and a Hermes plugin that can shadow or enable routing, skill suggestion, memory filtering, compaction selection, triage, mailbox sorting, and computer or browser action Choice. README documents sub-second latencies and sub-cent costs per decision class. Distinct from fast-jev-compaction (Claude-only context trim), hermes-jev-approvals (shell auxiliary), and keeltrace/hermes-jev (async supervision).

Problem: Hermes and IDE agents burn frontier tokens on routing, memory hygiene, skill pick, and GUI steps that are decisions, not essays.

Built for: Hermes Agent operators plus Claude Code and Codex users installing the kerpopule skill pack.

First indexed on Jev Directory: 2026-09-21

Creator and team

Name
kerpopule
Handle
@kerpopule
“Jev routing, memory, compaction, skill pick, and computer use for Hermes, Claude Code, and Codex.”

How Jev is used

Role in the product flow
Multi-skill decision layer: routing, retrieval, compaction, skill pick, triage, and GUI or browser Choice
Primitives
ChoiceScoreNoul
State in
Redacted user turns, passage batches, skill name lists, inbox messages, or safe action tables per skill README privacy rules.
Decision out
Model tier Choice, keep or drop turns, ranked passages, skill id or none, triage labels, next GUI or browser action with confidence.
  1. python3 install.py registers Hermes, Claude Code, and Codex when present
  2. jev setup-key stores TypeSafe credentials outside chat
  3. /jev routing shadow logs decisions before switching models
  4. Plugin tools jev_memory_filter, jev_compact_select, jev_choose_action expose hot paths

Each skill sends only the minimum redacted state Jev needs for a typed question: never full transcripts on routing, never screenshots on computer use, local injection screens before memory calls. Shadow mode lets operators compare Jev routing logs against incumbent models without switching production traffic. Computer and browser skills mirror the directory pattern of finite safe actions with Jev Choice per step. Compaction and handoff skills use parallel Score or keep or drop nouls instead of asking a frontier model to summarize tool noise. If you only need Claude Code context trimming, use tamaratran-fast-jev-compaction; if you need Hermes shell APPROVE gates, use anpicasso-hermes-jev-approvals. This repo is the broad everyday skill router Muhammad Aayan placed ninth on his September 2026 list.

Sourced performance claims

  • Model routing about 0.4 s per turn; skill selection across 377 skills in about 2.8 s per README table.Source: github.com/kerpopule/hermes-jev-skills README
  • Compaction eval: 58.7% recall alone and 75.0% with one search vs 37.5% and 68.3% baseline per SCORECARD-2026-09-20.md.Source: kerpopule/hermes-jev-skills evals/compaction

Features and stack

Features

  • Nine skills as portable SKILL.md files
  • Hermes plugin with shadow routing and dashboard
  • jev CLI for mail sorting, doctor, and model pool setup
  • Documented redaction and private profile rules
  • Featured in Aayan top-ten Jev repos thread

Stack

  • Python 3.9+ installer
  • Hermes Agent plugin API
  • TypeSafe System One

Pricing: Open source; TypeSafe keys via jev setup-key. README lists per-skill approximate dollar costs per thousand operations.

Demo

Showcase clip with the same lightbox player as the homepage. Click to play.

Muhammad Aayan@socialwithaayan

10 Jev repos blowing up right now

No video, just receipts: Aayan ranks ten GitHub projects by stars and tells you what Jev actually does in each loop. Bookmark the directory guide so you do not paste the same list into Slack twice.

guidesroundupinteresting

FAQ

Is this the same as fast-jev-compaction?
No. tamaratran/fast-jev-compaction is a Claude Code plugin for tool-row keep or drop. hermes-jev-skills is a multi-skill Hermes plus IDE pack with routing, memory, GUI, and more.
How do I try routing safely?
Start with /jev routing shadow on Hermes. It logs Jev decisions without switching models until you trust the pools.

How this differs from other Hermes plus Jev projects

fast-jev-compaction targets Claude Code tool rows. hermes-jev-approvals swaps the smart shell reviewer. keeltrace/hermes-jev adds async supervision. hermes-jev-skills is the day-to-day skill pack for routing, retrieval hygiene, and GUI steps with documented latencies.

Related learn guides

Original Jev guidance that pairs with this product pattern.

  • Jev use cases

    The patterns builders actually search for: moderation, routing, triage, RAG verify, and agent gates.

Related products

Hand-picked neighbors with rich profiles or overlapping tags.

Agent toolingTypeScript2,735
Featured
Claude Code plugin that replaces the compaction summary with Jev decisions: every tool call and result is scored in one fast request, stale ones are dro
agent toolingcompactionagents
Agent toolingPython3
PoC: TypeSafe Jev as the reviewer for Hermes Agent smart command approvals.
agent toolingagents
Agent toolingPython3
Typed System One decisions, ranking, verification, and an opt-in Hermes tool gate using TypeSafe Jev.
agent toolingagents
Official205
OfficialDemoFeatured
Official Claude Code / Codex skill pack for Jev primitives, patterns, and evaluation shape.
officialagentseval
By Rishit