AI
STATUS: LOGGED
QuickCmd
QuickCmd is my response to my own forgetfulness. I can never remember complex ffmpeg or docker flags, so I built an assistant that does it for me. But unlike other "AI Terminal" tools, this one is built on the principle of "Trust but Verify."
Core Stack
#Go#AI#CLI#Automation
The Intent
Copy-pasting AI commands from a browser into a terminal is dangerous. There needs to be a middle layer that vets the command before execution.
System Schematic
graph LR
A[User Note] -->|interpreted by| B(QuickCmd)
B -->|query| C[LLM API]
C -->|suggested cmd| D{Approval Prompt}
D -->|Verified| E[Execution]
D -->|Decline| F[Refine]
E -->|Optional| G[Docker Sandbox]
The Build
A Go binary that pipes your intent to an LLM, parses the response, and presents it in a clear "Accept/Decline" interface. Optionally runs in a container.
Reflections
I learned that parsing unstructured AI output is much harder than it looks. I had to build a robust state machine to handle weird formatting.