Skip to content

CONCEPT Cited by 1 source

Coverage cell

An (area × attack-class) matrix cell used to measure and track vulnerability scanning completeness across a codebase.

Definition

Each repository is decomposed into a grid where: - Area: a distinct subsystem, module, or code region within the repo - Attack class: a category of vulnerability (injection, memory corruption, timing side-channel, etc.)

Each cell represents "has this attack class been adequately tested against this area?" The Gapfill agent iteratively generates new hunt tasks for empty or under-tested cells until it stops producing findings.

Measurement

When prompts are updated, they are tested against a held-out repository to verify the total coverage-cell count actually moves — preventing prompt changes that reduce coverage.

Cost dynamics

Each additional gapfill pass costs roughly half as much as the initial hunt, making Gapfill the primary cost-to-coverage lever.

Seen in

Last updated · 542 distilled / 1,571 read