jsmastery-pro/skills

What it solves

It provides a structured, file-based engineering workflow for AI coding agents, preventing the loss of context that typically occurs in chat sessions. By moving the state of a project—such as scope, specifications, and context files—into the repository itself, it allows work to persist across sessions and be shared across a team of developers and agents.

How it works

The project implements a series of "skills" (modular instructions) that an AI agent can execute to move a feature from an idea to production. Instead of relying on chat history, each skill reads from and writes to specific files in the repository (e.g., AGENTS.md for context, docs/specs/ for architecture).

The workflow follows a logical progression:

1. Planning: /scope defines the product idea and /audit seeds the project context.
2. Design: /architect creates build specifications for load-bearing decisions.
3. Implementation: /develop builds the feature based on the spec, gating the process if a design decision is missing.
4. Verification: /check verify runs the app, /test writes test suites, and /check review performs a cross-model code review.
5. Finalization: /document generates PRs and changelogs, and /sync updates the project state.

Who it’s for

Developers using AI coding agents (such as Claude Code, Cursor, Codex, or Gemini CLI) who want a professional, disciplined software engineering process rather than ad-hoc prompting.

Highlights

• File-based State: Work survives session restarts because state is stored in the repo, not the chat.
• Defense-in-Depth Gating: Prevents implementation before design by forcing /develop to route back to /architect if a spec is missing.
• Workflow Depths: Offers suggested levels of rigor (Prototype, Alpha, Beta, GA) depending on the project's needs.
• Cross-Model Review: Supports using a different AI model for the /check review phase to ensure higher code quality.
• Monorepo Support: Scopes workflows to specific workspaces, each with its own context and commands.