Agent Skills

A collection of Agent Skills for Claude Code, Codex, and other compatible agents. Each skill lives in skills/<name>/SKILL.md, with optional references/ and other supporting files alongside it.

Available Skills

Skill	Description
brand-asset-audit	Audit and generate brand assets (logos, favicons, OG images, design system docs)
comprehensive-docs	Build a full `docs/` suite — index, categorized guides, role-based navigation, consistent formatting
deep-bug-hunt	Multi-agent bug-hunt expedition: lens-based finders, adversarial verification, dedup, and filing of verified defects
readme	Create or restructure a project's root README with the standard section layout and emoji-prefixed headers
dependabot-setup	Configure Dependabot on one repo, audit and fix an existing repo, or sweep a whole fleet
repo-ci-setup	SHA-pinned workflows and required checks that block PRs
homelab-github-settings	Adopt live SimplicityGuy repo settings (rulesets, labels, merge methods) into the homelab OpenTofu github module
sudo-l7	Staff-engineer working mode: check the premise, surface unmentioned risks, push back, verify, report truthfully

Install

This repo is both a Claude Code plugin marketplace and a Codex plugin marketplace, serving one plugin, agent-skills, that contains every skill.

Claude Code

claude plugin marketplace add SimplicityGuy/agent-skills
claude plugin install agent-skills@simplicityguy

Or inside a session: /plugin marketplace add SimplicityGuy/agent-skills, then /plugin install agent-skills@simplicityguy. Skills are namespaced by plugin, e.g. /agent-skills:readme.

Codex

codex plugin marketplace add SimplicityGuy/agent-skills

Then install agent-skills from the Plugins directory. Update later with codex plugin marketplace upgrade simplicityguy.

Standalone skills (no plugin)

Any tool that follows the Agent Skills layout can load the skill folders directly:

Run from the root of a clone of this repo:

# Claude Code (user-level)
mkdir -p ~/.claude/skills
ln -s "$PWD"/skills/* ~/.claude/skills/

# Codex (user-level)
mkdir -p ~/.agents/skills
ln -s "$PWD"/skills/* ~/.agents/skills/

For a single project, link into that project's .claude/skills/ or .agents/skills/ instead.

Adding a New Skill

Create skills/your-skill-name/SKILL.md (the directory name must match name)
Include YAML frontmatter with name and description fields
The description controls when the skill triggers -- include trigger keywords
Put long material in references/ and link to it from SKILL.md

Template:

---
name: your-skill-name
description: Use when [trigger conditions]. Triggers on "keyword1", "keyword2".
---

# Skill Title

## Overview
What this skill does.

## When to Use
- Condition 1
- Condition 2

## Process
Step-by-step instructions for Claude to follow.

CI

The Skill Lint workflow (.github/workflows/skill-lint.yml) runs two linters over skills/ on every pull request and on pushes to main:

agnix (agnix job), configured for Claude Code
skill-check (skill-check job)

The Skill Security workflow (.github/workflows/skill-security.yml) scans skills/ with two static security scanners and uploads SARIF to Code Scanning:

Cisco skill-scanner (skill-scanner job), failing on HIGH or above
NVIDIA SkillSpector (skillspector job), failing when the risk score exceeds 50

Both run without the LLM judge, so no secrets are needed. Accepted skillspector findings are recorded with reasons in .skillspector-baseline.yaml.

Both lint jobs are required status checks on main via a protect-main ruleset, managed in the homelab OpenTofu config. Changes to main go through a pull request.

License

See LICENSE.