Code Context
barkcli context benchmark
Measure the cold-start context payload (listing → keyword search → reading candidates) against one read of project_context.md. Deterministic proxy, no LLM — re-runnable on any repo in seconds.
Usage
barkcli context benchmark [options]Examples
barkcli context benchmarkbarkcli context benchmark --hits 3barkcli context benchmark --markdownbarkcli context benchmark --jsonFlags
--hitsCandidate files the reading step opens (default: 8)--linesLines read per candidate (default: 120)--jsonMachine-readable output--markdownCopy-paste claim block with the methodology footnote