Перейти к содержимому
TakoAPI
Назад к навыкам

wreckit-ralph

от christiancattaneo

Web & Frontend Development
413Число загрузок по данным clawskills.sh
0Просмотры
0 лайковВойдите, чтобы поставить лайк навыку
—Ещё нет оценок
Ваша оценка

Кратко

Bulletproof AI code verification.

Описание

Что делает этот навык

A multi-gate code verification agent that separates builder, tester, and breaker roles across independent parallel workers. Runs slop scans, type checks, mutation tests, and SAST analysis before aggregating results into a proof bundle with a Ship, Caution, or Blocked verdict. Supports TypeScript, Python, Rust, Go, Swift, Java, and Shell.

Splits verification into independent roles so the same AI context cannot confirm its own work, reducing silent failures from self-agreement.

Когда его использовать

  • Verifying AI-generated code before merging
  • Auditing a codebase for quality and coverage gaps
  • Fixing a bug and proving no regressions
  • Migrating a project to a new language or framework
  • Checking if a test suite actually catches real bugs

Пример сценария

ВВОД

User asks: Use wreckit to audit this Python project

АГЕНТ
  1. 1

    Runs detect-stack.sh to identify the language, framework, and test runner

  2. 2

    Runs slop-scan.sh to find placeholder code, empty stubs, and template artifacts

  3. 3

    Runs type checker and coverage stats, then spawns parallel gate workers for mutation testing and SAST

  4. 4

    Runs design-review.sh to analyze dependency graph, coupling, and circular dependencies

  5. 5

    Aggregates all gate results via proof-bundle.sh and writes .wreckit/decision.md with a final verdict

ВЫВОД

.wreckit/decision.md with a Ship, Caution, or Blocked verdict plus a full proof bundle at .wreckit/proof.json

Установка

clawhub install christiancattaneo/wreckit-ralph

Выполните команду CLI или вставьте промпт в Claude Code, Cursor, Codex — любой кодинг-агент, которым пользуетесь.

Ссылки