← All IntelClip / OtherDeferred tools & capping skill descriptions at 2% of context
From Codex, Behind the Harness — Dominik Kundel, OpenAI · ≈5:59
“we're marking some of these tools as deferred, and that means that they're not added directly to the context window, but instead are available through tool search later on.”
“for available skills, we actually cap the available skills list at 2% of your context total like maximum context window.”
What’s in it
- Explains how agents manage context bloat from many installed MCPs
- Reveals a 'deferred tools' trick to keep tool lists out of context
- Shows how skill descriptions auto-shrink to fit a 2% context cap
Clip transcript
context that is like growing as you're installing more MCPs. For that reason, we do two things to maintain cost. The first one is or the size of the context. The first one is having deferred tools. So, we're marking some of these tools as deferred, and that means that they're not added directly to the context window, but instead are available through tool search later on. Um and then the other part is that for available skills, we actually cap the available skills list at 2% of your context total like maximum context window. And that means that if it gets longer, we're reducing slowly the amount of like description that we're putting in there.
Comments
Checking sign-in…
Loading comments…