A practical guide to choosing between user prompts, system prompts, scoped instructions, Skills, Tools, Steering, and Delegation.
Thanks for writing this up. Scope and time are better axes than the usual list of what goes in the prompt, though the load-on-demand case is the awkward one, deciding to load is itself a model call you have to pay for.
I turned these concepts into a runnable tutorial using Python, now with 500+ GitHub stars: https://github.com/hardness1020/awesome-agent-architecture/tree/main/sections/10-system-prompt
Thanks for writing this up. Scope and time are better axes than the usual list of what goes in the prompt, though the load-on-demand case is the awkward one, deciding to load is itself a model call you have to pay for.
I turned these concepts into a runnable tutorial using Python, now with 500+ GitHub stars: https://github.com/hardness1020/awesome-agent-architecture/tree/main/sections/10-system-prompt