Browse / General-purpose / claude-workspace

claude-workspace

mauroepce

A senior workflow for Claude Code, encoded as 16 skills, 4 hooks, and one habit: open the session one level above the repos.

Open mauroepce/claude-workspace → Checked against the four rules.

ai-agentsclaudeclaude-codetemplateworkspace

Built for one agent

ships CLAUDE.md and nothing for other agents.

In its own words

README.mdmauroepce/claude-workspace

Mauricio's personal toolkit for working with Claude Code at senior level.

16 skills + a curated templates library, installable into your ~/.claude/ with one curl command. Layer on top of any project — new or existing, personal or team-owned — without touching the project's repo.

Each skill works two ways: invoke it manually (/work, /conventions, ...) exactly like the slash commands they evolved from, or let Claude auto-activate it — Claude reads each skill's description and applies the matching one when your request calls for it, without being asked. The methodology applies itself even when you forget to invoke it.

What it is

A codified senior workflow with five core principles:

  • Spec before generation — every task starts with a structured spec, not "hey claude write me X"
  • Critical review of agent output — never accept a first response, treat output like a junior's PR
  • Persistent context — conventions, architecture, decisions, todos all live in version-controlled files, not in chat history
  • Workflow as artifact — when a prompt repeats 3 times, it becomes a slash command
  • Atomic commands compose — small commands you can invoke standalone, or chain into orchestrators like /safe-commit

Install

Option A — as a plugin (recommended for new installs)

Inside any Claude Code session:

Quoting 26% of the README. Read the rest on GitHub ↗

Read from the repository's git tree on 2026-08-17.