Pre‑Change Complexity Mapping with CXCAP

Описание и применение

CXCAP – локальный инструмент статического анализа, который позволяет разработчикам видеть, как изменения, генерируемые ИИ‑агентами, влияют на сложность репозитория, выявляя горячие точки и зависимости до выполнения кода.

Задача

AI coding agents often pass tests but can silently increase repository complexity, making future changes harder.

Результат

CXCAP provides deterministic, local insight into repository complexity, helping developers avoid hidden debt that AI agents can introduce.

Рабочий процесс

• 1. Run cxcap audit . to map overall complexity and hotspots.
• 2. Run cxcap audit . --intent "<task description>" to discover likely touchpoints for a specific change.
• 3. Run cxcap audit . --focus <path> to see how a target file or component is connected to the rest of the repo.
• 4. Review the verdict (LOW/MODERATE/HIGH/SEVERE) and decide whether to proceed, refactor, or add tests.
• 5. If using an AI agent, feed the JSON report (--json) as structured evidence before generating code.

Доказательства

cxcap audit . --intent "change currency rounding" returned 3 likely touchpoints and a reasoning surface of 15 files across 5 components.

Источник

cxcap audit . --focus core/money.py showed that one file is imported by 23 outside files and reaches 19 further files transitively – 42 files exposed.

Источник

Historical --intent validation on 31 completed changes across 7 repositories scored Recall@10 0.70 and MRR 0.47.

Источник