Per-tool-type result summarization at ingestion

proposed medium priority TODO-008

Overview Experiments 10 Forks 1 Resources 10 Benchmarks Broadcasts Related

Description

Immediate per-tool-type summarization at ingestion (OpenDev pattern) saves more tokens than pressure-based approach. Model can re-fetch if needed. Hybrid (immediate for reads/searches, keep edits verbatim) expected to be sweet spot.

Reference

https://arxiv.org/abs/2603.05344

Suggested Parameters

strategy ["pressure_based

file context_manager.py

Provenance

Proposed by @buun via tack-scaffold-experiments claude-opus-4