Agent Skills Collection

A curated collection of reusable coding-agent skills for design quality, clean code, SOLID reasoning, and maintainability review.

This repository is structured to work well with:

• skills.sh
• skill-style repositories such as vercel-labs/agent-skills
• local Codex skill installation
• Claude Code project memory via root routing in CLAUDE.md and AGENTS.md

What This Repository Provides

• Focused skills with clear boundaries and minimal overlap
• Codex-facing metadata and Claude-facing adapters
• Reference material for progressive loading instead of one giant prompt
• Evaluation assets such as starter benchmarks, prompt packs, and scoring rubrics

Skill Catalog

Skill	Primary Use	Best For
apply-software-patterns	Pattern and architecture judgment	Deciding whether a pattern fits, comparing alternatives, improving boundaries, architecture direction
apply-clean-code	Readability-first refactoring	Naming, function clarity, practical cleanup, reducing incidental complexity
apply-solid-principles	SOLID-oriented design improvement	SRP, OCP, LSP, ISP, DIP, dependency direction, interface shaping
review-code-smells	Maintainability and code health review	Technical debt review, smell prioritization, hidden complexity, remediation planning

When To Use Which Skill

• Use skills/apply-software-patterns/SKILL.md when the question is about patterns, boundaries, architecture, or overengineering risk.
• Use skills/apply-clean-code/SKILL.md when the code works but is difficult to read, follow, or maintain.
• Use skills/apply-solid-principles/SKILL.md when the design pressure is about responsibility fit, extension points, substitution safety, or dependency direction.
• Use skills/review-code-smells/SKILL.md when the task is to review, prioritize, and explain maintainability smells.

Installation

skills.sh

Install a skill from this repository with:

npx skills add namikazebadri/basic-programming-skill --skill <skill-name>

Examples:

npx skills add namikazebadri/basic-programming-skill
npx skills add namikazebadri/basic-programming-skill --skill apply-software-patterns
npx skills add namikazebadri/basic-programming-skill --skill apply-clean-code
npx skills add namikazebadri/basic-programming-skill --skill apply-solid-principles
npx skills add namikazebadri/basic-programming-skill --skill review-code-smells

Install from a local checkout:

npx skills add . --skill apply-clean-code

Codex

Copy one skill folder into your Codex skills directory:

mkdir -p "$CODEX_HOME/skills"
cp -R skills/apply-clean-code "$CODEX_HOME/skills/"

Repeat for any additional skill you want installed.

Example:

cp -R skills/apply-software-patterns "$CODEX_HOME/skills/"
cp -R skills/apply-solid-principles "$CODEX_HOME/skills/"

Claude Code

This repository includes a root multi-skill router in:

• CLAUDE.md
• AGENTS.md

Claude Code can use these files to route tasks to the most relevant skill adapter under skills/<skill-name>/agents/claude.md.

Repository

• GitHub: namikazebadri/basic-programming-skill

Repository Layout

skills/
  apply-clean-code/
    SKILL.md
    agents/
    references/
    scripts/
  apply-solid-principles/
    SKILL.md
    agents/
    references/
    scripts/
  apply-software-patterns/
    SKILL.md
    agents/
    references/
    scripts/
  review-code-smells/
    SKILL.md
    agents/
    references/
    scripts/
CLAUDE.md
AGENTS.md
README.md

Evaluation Support

Several skills in this repository include:

• starter benchmark packs for quick Codex vs Claude comparison
• full eval prompt sets
• evaluation rubrics
• benchmark init and summary scripts

Examples:

• skills/apply-software-patterns/references/benchmark-harness.md
• skills/apply-clean-code/references/benchmark-harness.md
• skills/apply-solid-principles/references/benchmark-harness.md
• skills/review-code-smells/references/benchmark-harness.md

Design Philosophy

These skills are designed to be:

• problem-first instead of buzzword-first
• incremental instead of rewrite-happy
• readability-aware instead of abstraction-heavy
• practical enough for real codebases under delivery pressure

Notes

• Not every skill should be loaded for every task.
• The root router is intentionally lightweight so Claude can choose relevant skill context instead of loading the entire repository.
• The benchmark assets are intended to help compare agent behavior, not to force one rigid style.