Menu
Enforces build, test, and lint gates with a self-correction loop and retry budget.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Scans the codebase for tech debt — balls of mud, shallow modules, and DRY violations — then presents prioritized bottlenecks and hands off to TDD-driven refactoring via superpowers.
Forces DDD ubiquitous language definition before writing complex logic. Grills the user on business terminology, extracts a glossary, and documents it before handing off to brainstorming.
Sets up a new project for the AI workflow — creates AGENTS.md, mise.toml tasks, initializes CodeGraph, and seeds Engram memory. Optional/auxiliary skill.
Evidence-first development for high-stakes code. Writes a concrete spec, enforces strict TDD, then runs code through a gauntlet of 6 checks and generates an EVIDENCE.md report with raw numbers.
Enforces Red-Green-Refactor Test Driven Development workflow.
Manages Architecture Decision Records (ADRs) when modifying architecturally significant files.
| name | verification-gates |
| description | Enforces build, test, and lint gates with a self-correction loop and retry budget. |
To ensure repository stability, all major changes must pass through a series of "Verification Gates."
npm run build (or equivalent).npm test (or equivalent).npm run lint or ruff check.Whenever a gate fails (e.g., via the gate tool), follow this loop:
stderr and stdout provided by the gate tool.gate tool with the same command.Always run the gate tool for any verification step. If a gate fails, announce "SELF-CORRECTION LOOP START [Attempt X/5]".