skills.rest
GreenyZA
GreenyZACOMMUNITY
Explore all skills by this author →
GitHub

light

Enforces Light Engine identity and local-first model routing rules every session turn.

Updated Sep 20, 2026
Search Tags:#model routing#openrouter#vulkan#ollama#free tier#agent identity#local llm
One-click install
npx skills add https://github.com/GreenyZA/neo-light --skill light-greenyza
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: light
Source: https://github.com/GreenyZA/neo-light/tree/main/.grok-home/skills/light
Command: npx skills add https://github.com/GreenyZA/neo-light --skill light-greenyza

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? This Skill prevents identity confusion and wasted spend by keeping the agent locked to the Light Engine persona and its local/free-first model routing order instead of drifting to paid premium models. ## Core Features & Use Cases - Always-on identity enforcement: Establishes the Light Engine persona (WarPuppy, English, short answers) on every turn in the session. - Model routing order: Enforces the fallback chain vulkan-3b → ollama-7b → minimax-m3 → openrouter-free, blocking SuperGrok usage in this environment. - Environment guardrails: Restricts vault writes to 01-Projects/Light-Engine/, blocks mail/calendar/GoDaddy/public git actions, and provides an escape hatch via /home/warpuppy/bin/back-to-james when the user wants James back. - Use Case: A user running a lightweight local-GPU agent session activates this Skill so the agent never escalates to paid models and stays within the Light Engine workspace boundaries. ## Quick Start Activate the light skill at the start of a Light Engine session to enforce local-first model routing and identity rules.

Frequently Asked Questions about light

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I keep an AI agent on local and free models only?▼

Use an always-on routing skill that defines a fixed model order such as vulkan-3b, ollama-7b, minimax-m3, then openrouter-free. The agent checks this chain every turn and stops instead of calling paid models like SuperGrok.

What is the model fallback order in Light Engine?▼

Light Engine tries vulkan-3b first, then ollama-7b, then minimax-m3, and finally openrouter-free models. SuperGrok is treated as the wrong environment and the agent stops and reports the correct environment instead.

Can I switch from Light Engine back to the James agent?▼

Yes, say you want James back and the agent runs /home/warpuppy/bin/back-to-james immediately without waiting. This hands the session off to the heavyweight agent environment.

What context window does the Vulkan local model support?▼

The Vulkan setup in this skill is limited to 4096 tokens of context. Requesting 64k context is not supported on this card; that capability belongs to the Hermes environment instead.

What actions are blocked in a Light Engine session?▼

Light Engine blocks mail and calendar writes, GoDaddy operations, public git pushes, and interactions with Olivia. Vault notes may only be written under 01-Projects/Light-Engine/ or the current repository.

Related Skills

Explore More Software Engineering Skills→
AveerAveer
1

zeus-llm

Orchestrate deployment and configuration of local LLM inference servers for OpenCode.

Community
Intermediate
TechNickAITechNickAI
1

cron-healthcheck

Detect failing Hermes cron jobs and escalate remediation tasks hourly.

Community
Intermediate
bennybennisonbennybennison

cost-tracking

Track token usage and estimate costs for conversational AI sessions.

Community
Intermediate
cianos95-devcianos95-dev

context-window-management

Enforce a 3-tier delegation model and context budget controls for AI agents.

Community
Advanced