{'>'}_
English 繁體中文 日本語 한국어 Deutsch Español Français
Login

Structured development workflow that auto-routes coding tasks to the right execution mode — quick fixes, multi-step plans, research, shipping, or verification. Enforces quality gates (lint, types, tests), multi-stage code review, and atomic commits. Use when starting a feature, fixing a bug, refactoring code, planning implementation, shipping a release, or running quality checks.

10 stars
1 forks
JavaScript
114 views
View on GitHub Add to Favorites

SKILL.md

name: mint description: "Structured development workflow that auto-routes coding tasks to the right execution mode — quick fixes, multi-step plans, research, shipping, or verification. Enforces quality gates (lint, types, tests), multi-stage code review, and atomic commits. Use when starting a feature, fixing a bug, refactoring code, planning implementation, shipping a release, or running quality checks."

mint

Disciplined agentic development — fresh context per task, zero slop. Routes every coding task to the right execution mode, enforces quality gates, and manages multi-stage review pipelines.

Quick Start

Invoke mint via the Skill tool with a task description. mint auto-detects the right mode:

Mode When What it does
quick Task touches ≤3 files, scope obvious Direct implementation with quality gates
plan Multi-file changes, ambiguous scope Decomposes into specs, implements in phases
ship Multiple features, "build all" Parallel spec execution with coordinated review
research "How to", "compare", investigation Structured research with documented findings
verify "Check gates", "audit" Runs lint, types, tests and reports status

Routing Logic

The full orchestrator at skills/mint/SKILL.md evaluates signals in order — first match wins. Override with "just quick-fix it" or "plan this out" to force a mode.

What mint Enforces

  • Quality gates — lint + types + tests before every commit
  • Multi-stage review — spec review, then parallel audit (quality, security, conventions, tests)
  • Atomic commits — one commit per logical change, never push (human reviews and pushes)
  • Context protection — heavy work delegated to subagents, main context stays clean

For the auto-trigger rule (invoke mint before any file modification), see skills/using-mint/SKILL.md.