AbsolutelySkilled: Install 156 Production-Ready Skills to Make Claude Code Remember Your Conventions

A new registry of structured skill modules (SKILL.md files) lets you install persistent domain expertise into Claude Code once, guiding its behavior across all future sessions.

GAla Smith & AI Research Desk·Mar 16, 2026·4 min read·71 views·AI-Generated

Source: reddit.comvia reddit_claude, gn_claude_api, hn_claude_code, reddit_claude, medium_claude, gn_claude_code_tips, hn_claude_cli, gn_claude_codeWidely Reported

What It Does — A Registry of Skills for Persistent Knowledge

AbsolutelySkilled is a GitHub registry containing 156 structured skill modules you can install into Claude Code. Each skill is a SKILL.md file designed to solve a core frustration: Claude Code's lack of persistent knowledge between sessions. Instead of re-explaining your project's conventions, architecture, or preferred patterns every time you start a new chat or project, you install a skill once. When triggered, Claude Code loads the structured knowledge from the skill file, guiding its behavior consistently.

These aren't just simple prompts. Each skill includes:

Trigger Conditions: When and how the skill should activate.
Reference Files: Supporting documentation or examples.
Evals: 10-15 test cases to verify the skill works correctly.
Anti-patterns: Explicit guidance on what not to do.

The registry covers a wide range of domains crucial for developers, including system design, Docker, Kubernetes, React, Next.js, PostgreSQL, security review, technical writing, and SEO.

How It Works — Structured Skills as SKILL.md Files

The system works by extending the concept of a CLAUDE.md file—a project-level guide for Claude Code—into reusable, installable modules. You install skills via a command-line tool (npx skills). Once installed, these SKILL.md files sit in your project or a global directory. When Claude Code encounters a task that matches a skill's trigger conditions, it loads the relevant SKILL.md file into its context, applying the pre-defined expertise and rules.

This creates a form of persistent, cross-session memory. The knowledge isn't stored in a model's weights but in accessible, version-controlled files that Claude Code can reference on-demand.

Two Standout Skills: Superhuman and Second Brain

The creator highlights two foundational skills that reimagine the development workflow for AI constraints:

1. Superhuman
This skill rearchitects the entire development lifecycle into a dependency-aware, parallelized process. It doesn't just write code; it manages complex tasks like a project manager. Its 7-phase workflow is:

INTAKE: Understand the task.
DECOMPOSE: Break it into a dependency graph (DAG) of sub-tasks.
DISCOVER: Gather necessary context and files.
PLAN: Create an execution strategy.
EXECUTE: Run independent tasks in parallel using sub-agents.
VERIFY: Enforce Test-Driven Development (TDD) at every step with verification loops.
CONVERGE: Integrate results.

It maintains a persistent board.md file that survives across sessions and context resets, acting as a project dashboard.

2. Second Brain
This skill creates a persistent, tag-indexed memory system for your agent. It writes learnings to a ~/.memory/ directory that survives across all projects and sessions. To maintain context efficiency, it enforces a 100-line ceiling per memory file and uses wiki-style links for graph navigation. After complex tasks, it auto-proposes new entries for this memory.

Try It Now — Installation and First Steps

Getting started is straightforward using npm/npx:

Install a single skill (e.g., the superhuman workflow):

npx skills add AbsolutelySkilled/AbsolutelySkilled --skill superhuman

Install all 156 skills at once:

npx skills add AbsolutelySkilled/AbsolutelySkilled -g

After installation, the skills will be available in your configured skills directory. When you start Claude Code in a project, it will begin to recognize and utilize the installed skills based on their triggers. Start with a focused skill like react-best-practices or docker-compose for your current work, then explore the superhuman skill for larger, multi-file refactors or feature builds.

Why This Changes Your Workflow

This registry directly tackles the "amnesiac agent" problem. By externalizing expertise into installable, version-controlled skill files, you stop losing time on repetitive explanations. Your team can share a core set of skills, ensuring consistency. The superhuman skill, in particular, introduces a structured, parallelized approach to development that leverages Claude Code's ability to manage multiple threads of thought, potentially cutting down the time for large tasks significantly.

Following this story?

Get a weekly digest with AI predictions, trends, and analysis — free.

AI Analysis

Claude Code users should immediately view this as a system for creating and sharing persistent team knowledge. Instead of pasting the same architecture notes into every new chat, package them as a skill. Start by installing the `superhuman` skill for your next medium-complexity task (like adding a new API endpoint). Observe how it decomposes the work and manages the `board.md`. This forces a more systematic approach. Next, audit your last week of Claude Code sessions. What conventions did you explain more than once? Package those as your own local `SKILL.md` file using the registry's format as a template. This could be your team's ESLint rules, API response format, or error handling pattern. The goal is to shift from ad-hoc prompting to a library of reusable expertise. Finally, integrate the `second-brain` skill. Let it run for a few days as you work. Periodically review the `~/.memory/` directory it creates. This becomes a searchable knowledge base of lessons learned, which you can then refine into new, more precise skills. This turns Claude Code from a tool that forgets into one that accumulates wisdom.

#best-practices #cli #mcp #workflow

Enjoyed this article?

Get the weekly AI intelligence briefing