How-to · SEL · SEL.4

Scheduled tool-calls vs a product labeled 'agent'

A product labeled 'agent' may just be a scheduled prompt that calls tools toward a goal — test for schedule + tools + goal + human gate, not branding.

Open your notes and pick one tool that's marketed as an 'agent.' Score it Y/N on: (1) scheduled runs (2) tools beyond chat (3) an explicit goal (4) a human gate for send/spend. You're done when you've written the plain-English pattern in 10 words or less.

Worked example

Here's a useful question from Simon Willison: what's actually different between that flashy product and a simple cron-like schedule telling a model to call some tools? The workplace skill is opening the product card and scoring Y/N on schedule, tools, goal, and human send/spend. If it's really just cron plus chat, use it that way — with limits. You still own deploy and send.

← All How-tos