B-TIER technique
PromptOps Smart Action Detector — save tokens, run commands locally
100k ft — What is this?
PromptOps Smart Action Detector — save tokens, run commands locally
50k ft — Why care?
@shellonback on PromptOps' new detector: when a sub-agent is about to run a directly-executable command (npm run deploy/build/test), it suggests running in the terminal instead of sending to the LLM. Concrete anti-token-waste pattern.
10k ft — Key takeaways
• Hard rule: never ask an LLM to run a command you could bash yourself
5k ft — Tools & techniques
(no tools tagged)
AI agents, automation
1k ft — Deep-dive / try it
How we reduced token waste (caused by pure laziness)
How many of you do this? 👇
Typing commands into AI like:
npm run deploy
npm run build
npm test
…when you could run them in 2 seconds from the terminal.
We realized we were doing it too.
Simple commands, already defined in project config files, automation scripts, and build systems…
yet we were still sending them through AI → wasting tokens.
So in PromptOps we introduced a Smart Action Detector.
When you’re about to spawn a sub-agent:
• it understands what you want to do
• checks if the action is directly executable
• and if AI isn’t needed…
👉 it suggests running it directly in the terminal
If you confirm:
• it opens the terminal
• injects the command
• runs it instantly
🎥 In the video:
we run npm run deploy
and instead of starti…