Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Hermes System & DevOps Pack

    2

    Keep an AI workstation fast, free and quiet.

    $19

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Hermes System & DevOps Pack

    Hermes System & DevOps Pack

    Example session with this skill installed

    Launch OmniRoute silently, discover any new free AI models available, probe each one with a real chat completion, and append the working ones to my eco fallback combo. Then add NVIDIA NIM as a cloud vision fallback for the profiles that have no local vision backend.

    • Read your context and instructions
    • Compiled the hermes system

    OmniRoute is up (hidden, idempotent — already running, no second instance). I probed 5 newly-listed free models and 3 answered correctly, so they were appended to the eco combo. NVIDIA NIM is now registered as the vision fallback.

    System & DevOps Pack output

    Free model discovery

    ModelProbe resultAction
    meta-llama/llama-4-scout200 OK, 1.2sAdded to eco combo
    qwen/qwen-3-32b200 OK, 0.8sAdded to eco combo
    mistralai/mistral-small-3200 OK, 1.5sAdded to eco combo
    google/gemma-3-27btimeout 30sSkipped (dead)
    cohere/command-r-plus429 rate-limitSkipped (quota)

    Backup taken before write: backups/eco-2026-10-11T14-22-31.json

    Vision fallback

    
    # config.yaml (vision section)
    vision:
      provider: nvidia-nim
      model: meta/llama-3.2-90b-vision-instruct
      api_key_env: NVIDIA_API_KEY
      endpoint: https://integrate.api.nvidia.com/v1
    
    config.yaml

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Keep an AI workstation fast, free and quiet — without a paid model subscription and without babysitting Windows.

    What it does

    Four production-ready Hermes Agent skills that keep a local LLM stack running cleanly and cheaply:

    • omniroute-auto-update — discovers newly available free AI models, probes each one with a real chat completion, and appends only the models that actually answer to the end of your eco fallback combo (backup taken first). Also launches OmniRoute invisibly and idempotently.
    • nvidia-nim-vision — a cloud vision/LLM fallback via NVIDIA NIM when no local vision backend is available, plus synthetic-video detection and NVIDIA-hosted image generation — all on one free nvapi- key.
    • windows-optimization — audits a Windows desktop (power plan, GPU, RAM channel mode, disks/SMART, startup entries) and applies only the changes that are real levers: power/GPU tuning, debloat, temp/recycle-bin cleanup, startup hygiene. Refuses the well-known myths and flags the one genuine security-vs-performance tradeoff (HVCI) before touching it.
    • security-monitoring — tames a continuous log/port monitoring stack that relays alerts to Telegram via scheduled tasks: silences false-positive alert spam at the monitor's rule level, never by editing the Hermes core, and verifies that an exclusion did not mask a real crash or auth failure.

    Why use this pack

    Most LLM-stack guides assume an unlimited budget and a cloud-only setup. This pack assumes the opposite: a single workstation, free-tier models where they genuinely work, and a machine that stays healthy over months of use. Every command was validated on a real Windows build machine, and each skill documents its own pitfalls, provider quirks and known limits.

    The result is a workstation that boots quietly, routes around dead free models automatically, and tells you when something is actually wrong instead of flooding a chat with false alarms.

    How to use it

    1. Install Hermes Agent and drop the four skill folders into your skills directory.
    2. Copy each skill's .env.example to .env and fill in your values (OMNIROUTE_API_KEY for the routing skill, NVIDIA_API_KEY for the vision skill, optional Telegram variables for notifications).
    3. Ask your agent to run a skill, or follow the numbered procedure in each SKILL.md.

    Requirements

    • Hermes Agent v0.21+
    • Python 3.11+
    • PowerShell 5.1+ (Windows) for windows-optimization
    • Windows for the OmniRoute hidden launcher and the Windows optimization scripts; the NVIDIA NIM and monitoring skills are cross-platform
    • Optional: a free NVIDIA NIM key (nvapi-, created at build.nvidia.com)

    Security

    Scanned and cleared against the Agensi 8-point security review (8/8):

    • No secrets, no telemetry, no obfuscation, no prompt injection
    • The only in-code network call is the local OmniRoute gateway (http://127.0.0.1:20128) — no external host is contacted by the packaged scripts
    • NVIDIA endpoints are documented API calls the user makes with their own key
    • Environment variables read: OMNIROUTE_API_KEY, NVIDIA_API_KEY, and optional TELEGRAM_BOT_TOKEN / TELEGRAM_CHAT_ID

    License

    Proprietary — all rights reserved. Commercial license granted on purchase. A NOTICE.md file documents the license correction from an earlier template default.

    Contact

    Thomas Leroyer — searching.murphy@gmail.com — https://www.agensi.io/@thomas-leroyer

    • Windows uniquement pour windows-optimization et security-monitoring (PowerShell 5.1+, admin requis pour les points de restauration).
    • omniroute-auto-update attend un gateway OmniRoute local sur 127.0.0.1:20128 ; si absent, le script s'arrête proprement sans modifier le système.
    • nvidia-nim-vision nécessite une clé NVIDIA_API_KEY valide et un accès sortant vers integrate.api.nvidia.com / api.nvcf.nvidia.com / build.nvidia.com.
    • Les points de restauration et l'audit de démarrage ne couvrent pas les politiques de groupe d'entreprise (GPO) ni les postes gérés par MDM.
    • Aucun rollback automatique en cas d'échec d'une optimisation : un point de restauration est créé AVANT chaque changement, la restauration reste manuelle.
    • Pas encore testé sur Windows Server ; conçu et validé sur Windows 10/11 grand public.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    Requires

    Python
    Windows

    Permissions

    Terminal / Shell
    Network Access
    Read Files
    Write Files
    Environment Variables

    File Scopes

    **/SKILL.md
    **/references/*.md
    **/.env.example
    **/backups/*.json
    **/scripts/*.py
    **/scripts/*.ps1
    **/scripts/*.vbs

    Permissions: - Network: local gateway only (127.0.0.1:20128) from the packaged code; the vision skill documents user-initiated calls to NVIDIA NIM (integrate.api.nvidia.com, api.nvcf.nvidia.com, build.nvidia.com). - Files read: ~/.omniroute/.env (Bearer key), skill folders, .env.example. - Files written: backups/eco-<timestamp>.json (combo backup before any write), references/model_verdicts.md (dated probe log), temp/recycle-bin cleanup and system restore points (Windows optimization, with a restore point created before any change). - Env vars: OMNIROUTE_API_KEY, NVIDIA_API_KEY, optional TELEGRAM_BOT_TOKEN / TELEGRAM_CHAT_ID. - No hardcoded secrets, no telemetry, no obfuscation, no prompt injection. - Backup combo written before any write; Windows restore point created before any change.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    No. The routing skill is built around free-tier models and only adds a model to your combo after it passes a real inference probe. The NVIDIA NIM key is free (you create it at build.nvidia.com). Nothing in the pack requires a paid plan.

    Two of the four skills are Windows-specific — the OmniRoute hidden launcher uses wscript/netstat, and windows-optimization uses PowerShell. The NVIDIA NIM vision skill and the security-monitoring skill are cross-platform.

    Yes. One free nvapi- key works across the NIM catalog (vision LLM, image generation, synthetic-video detection). Free-tier large models have long cold starts, and image generation on a free key is unreliable — both limits are documented in the skill.

    Non. Aucun appel réseau sortant dans les scripts packagés, sauf : la lecture du gateway OmniRoute local (127.0.0.1:20128) — jamais en dehors de votre machine ; les appels NVIDIA NIM que vous initiez (vision), vers les endpoints integrate.api.nvidia.com / api.nvcf.nvidia.com / build.nvidia.com — déclarés dans les permissions. Aucune télémétrie, aucun tracking, aucun envoi de fichiers.

    Oui. windows-optimization crée un point de restauration Windows avant chaque modification. La restauration se fait via l'outil standard (rstrui.exe). Aucun rollback automatique n'est déclenché — c'est volontaire, pour que vous gardiez la main.

    Pas officiellement. Le pack est validé sur Windows 10 et 11 grand public. Sur Windows Server ou un poste avec GPO / MDM, certaines optimisations peuvent être bloquées ou ignorées — à tester dans votre environnement avant déploiement.

    New in DevOps & Cloud