Context Selection
Ranking garbage still packs garbage.
A librarian pulling three books, not wheeling the whole stack to your desk.
Visual Workflows
Start here — scroll inside each diagram frame to explore, then use + / − to zoom up to 200% if needed.
Overview
Scroll inside the frame to explore · use + / − to zoom up to 200%
Scroll inside the frame to explore · use + / − to zoom up to 200%
This turn
Scroll inside the frame to explore · use + / − to zoom up to 200%
A librarian pulling three books, not wheeling the whole stack to your desk.
Key Takeaways
- 1.Selection chooses which memories and chunks are even candidates.
- 2.Filter by metadata, recency, and permission before you rank.
- 3.Empty selection is allowed — do not invent hits.
- 4.Fewer, better pieces beat a dump of forty maybe-relevant rows.
- 5.Ranking garbage still packs garbage.
Real Example
Scenario
Selection: pack only what this turn needs.
What you would do
In Context Engineering, apply Context Selection to this scenario: Selection: pack only what this turn needs. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.
Commands
Commands to Remember
Selection chooses which memories and chunks are even candidatesFilter by metadata, recency, and permission before you rankEmpty selection is allowed — do not invent hitsFewer, better pieces beat a dump of forty maybe-relevant rows
Cheat Sheet
Quick recap
quick ref- •Selection
- •Budget
- •Don't dump
Common Mistakes
- ✕Skipping evaluation for Context Selection before production
- ✕No logging or tracing around context selection steps
- ✕Ignoring cost and latency implications