What problem does it solve? Personal automations that read messages and act on your behalf are risky: they can execute destructive commands, spam contacts, or leak data to third-party services. This Skill defines a safe architecture for building local assistants that run entirely on your machine with no paid SaaS dependencies. ## Core Features & Use Cases - Layered Pipeline Design: Separates listener, command parsing/validation, prompt building, local model calls, and response sending into single-purpose modules. - Safety Guardrails: Enforces command prefixes, sender allowlists, input validation, rate limiting, and a strict ban on shell execution or file deletion. - Local Model Integration: Connects to local LLM HTTP APIs (e.g. Ollama at localhost:11434) with timeout handling and graceful failure messages. - Use Case: Build a chat-triggered assistant where browser automation reads your messages, accepts only /ai ... commands from your own number, sends text to a local model, trims replies to ~1,000 characters, and logs each action to a local audit trail. ## Quick Start Use the local-first-ai-automation skill to design a safe local chat bot that answers /ai commands through a local Ollama model with allowlisted senders and no destructive actions.
