Review lenses

A lens is an adversarial angle of attack. The DELIVER reviewer does not read the code “in general”: it examines it 4 times, from 4 independent angles, then synthesises a weighted verdict.

Why — the problem it solves

A single reader has blind spots: they validate what they already know to look for. By breaking the review into independent lenses, each weakness is hunted for its own sake — by a lens whose sole mission it is. None may “let it slide” on behalf of the others.

Key concepts — how it works

Lens Angle of attack What it opposes
cold-reader Reads code and tests with zero prior context. Verifies business language, naming clarity, intent visibility.
architecture-boundaries Verifies the dependency direction of Clean Architecture. No mocks in Domain/Application, Object Calisthenics on the Domain.
test-integrity Hunts test theater and Iron Rule violations. Tests that assert nothing, tests coupled to implementation, fake RED/GREEN.
quality-gates Falsifies the quality-gates evidence log against the Git tree. Read-only: confronts what is declared with what is actually committed.

The 4 verdicts are weighted then synthesised (Genesis A7 pattern): a single BLOCKER on one lens is enough to reject.

Conditional fidelity lenses (DELIVER)

On top of the 4 CORE lenses, two fidelity lenses join the panel only when the matching capability is active. They enter when the software-engineer has delegated test wiring to a worker (see the DELIVER fan-out).

Lens Activated when What it opposes
mock-fidelity-lens mock-integration-worker wired a mock. Does the mock faithfully reflect the downstream contract (statuses, headers, error shapes) rather than a complacent double?
contract-fidelity-lens contract-testing-worker wired a contract test. Does the test actually cover the provider contract (schema, codes, ProblemDetails) without bypassing it?

They follow the same synthesis rule: a BLOCKER on a conditional lens rejects just like a CORE lens.

Why this practice

« The combined attention of several reviewers finds defects a single reader misses. » — Wiegers, K., Peer Reviews in Software, 2002.

Several independent lenses are the automated equivalent of several reviewers with complementary skills.

Pitfalls & anti-patterns

Detail pages per lens

Going further

Sources

Terms to know: lens, test theater, Iron Rule of tests, Object Calisthenics — see the glossary.