Give pass criteria instead of instructions, and lock procedures into skills

This introductory walkthrough recasts an agent known as a coding tool as a general-purpose office-automation tool to hand document, spreadsheet and image work to, checking its features one at a time on real tasks. The key is the distinction between working modes: in one mode, once the completion conditions are fixed up front, the agent repeats edits and checks on its own until it reaches that state, so the round trip of a person looking at the result and asking again moves into an internal loop. Also presented is a way to save a procedure that came out well once, as is, under a name and call it back with a single line; in practice the bigger point is that for jobs that run the same procedure every time, the cost of rewriting the instructions disappears, and skipping the planning step even reduces usage. It shows, with a task that moves a document into a spreadsheet, that even programs with no integration point can be hooked up through screen control. However, it draws the line itself, saying that this screen control is hard to use in real work for now because of its speed and usage, and any discussion of permissions or undo is missing.

Key points

  • Fixing completion conditions first moves the checking round trip into the agent's internal loop
  • Locks a procedure that worked into a skill and recalls it in one line, cutting usage too
  • Brings programs with no integration point into automation scope via screen control
  • Admits screen control is not yet practical for real work because of speed and usage

This is a note: my own write-up after reading several sources. It has no source link; events checked against a source are kept as records.

Same week's notes

Is it still holding the first rule at the tenth file?Agents meant to be isolated exchanged signals with each otherA design that doesn't hand regulatory review wholesale to one modelWhat held back the switch was accumulated setup, not performanceA flood of malicious packages in a short span, and the bar for isolation