Agent Skills

Open-source skills for AI coding agents. Works with Claude Code, Codex, Cursor, Windsurf, and any agent that reads SKILL.md files.

Some of these skills I found and kept. Some I built at 2 a.m. when nothing else fit. If it's here, it worked.

— Roman

Skills

Skill	Description
humanizer	Rewrite AI-sounding text so it reads like a real person wrote it. Removes inflated language, fake depth, vague authority, repetitive structure, and other common AI writing tells.
anthropic-managed-agents	Primer + setup coach for Anthropic Claude Managed Agents — the hosted agent runtime with tools, MCP connectors, and skills. Turns an idea or spec into a setup blueprint.
vibe-audit	Audit your vibe-coded app against 34 security, performance, and code quality checks. Generates a report with pass/fail scorecard and self-contained fix files.
moat-analyzer	Analyze any business against the 10 Moats framework. Scores LLM disruption risk, moat durability, and identifies startup opportunities.
marketing-skills-for-agents	34 marketing skills — CRO, copywriting, SEO, GEO, analytics, paid ads, email sequences, sales enablement, and growth engineering.

Every skill folder has its own README.md with full description and manual install fallback.

Install a single skill

humanizer

npx skills add https://github.com/BadTechBandit/skills --skill humanizer

anthropic-managed-agents

npx skills add https://github.com/BadTechBandit/skills --skill anthropic-managed-agents

vibe-audit

npx skills add https://github.com/BadTechBandit/skills --skill vibe-audit

moat-analyzer

npx skills add https://github.com/BadTechBandit/skills --skill moat-analyzer

marketing-skills-for-agents

npx skills add https://github.com/BadTechBandit/skills@marketing-skills-for-agents

Install the whole repo

npx skills add BadTechBandit/skills

Manual install (any skill)

git clone https://github.com/BadTechBandit/skills.git
ln -s $(pwd)/skills/<skill-name> ~/.claude/skills/<skill-name>
ln -s $(pwd)/skills/<skill-name> ~/.codex/skills/<skill-name>

How it works

Each skill is a folder with a SKILL.md file and optional references/, scripts/, and assets/ directories. Your coding agent reads SKILL.md and follows the instructions using whatever tools it has available.

vibe-audit/
├── README.md                         # Landing page + install
├── SKILL.md                          # Main instructions (what your agent reads)
└── references/
    ├── security.md                   # 10 checks
    ├── database-performance.md       # 4 checks
    ├── infrastructure.md             # 8 checks
    └── code-quality.md               # 3 checks

---

For agents: how to add a new skill to this repo

When Roman asks you to add a new skill to this repository, follow this exact workflow. No back-and-forth — everything you need is here.

1. Choose a skill name

• Lowercase, kebab-case (my-new-skill)
• Descriptive and specific — name matches the frontmatter name: field and the folder name
• Avoid personal/project branding; prefer generic names that describe the capability (e.g. anthropic-managed-agents, not roman-agents)

2. Create the skill folder

At the repo root:

skills/<skill-name>/
├── README.md          # Landing page for anyone who opens just this folder on GitHub
├── SKILL.md           # The actual skill (frontmatter + instructions your agent reads)
└── references/        # Optional: detail files loaded on demand
    └── *.md

Use scripts/ or assets/ subdirectories only when genuinely needed (executable code or output templates). Most skills don't need them.

3. Write the SKILL.md

Frontmatter must have exactly two fields:

---
name: <skill-name>
description: <what it does + explicit "use when" triggers — this is the only thing the agent reads to decide whether to load the skill>
---

Rules for the body:

• Imperative voice ("Do X", not "You should do X")
• Keep under ~500 lines — move detail into references/*.md
• No "when to use this" sections in the body (put that in frontmatter description)
• No README-style preamble, changelog, or setup instructions inside SKILL.md

4. Scrub PII

Before committing, check for personal names, company names, project names, emails, and internal product names. Replace with generic examples ("a SaaS product", "an internal tool"). Run:

grep -riE "roman|gotmoat|curo|flyworx|creworx|glitch" skills/<skill-name>/

Zero hits = safe to commit.

5. Write the per-skill README.md

Every skill needs its own README.md following this exact template:

# <Skill Display Name>

<One-paragraph description of what the skill does. Mirror the SKILL.md frontmatter but in human-friendly prose.>

**Use it when you want to:**
- <Concrete use case 1>
- <Concrete use case 2>
- <Concrete use case 3>

## Install

​```bash
npx skills add https://github.com/BadTechBandit/skills --skill <skill-name>
​```

Browse on skills.sh: https://skills.sh/BadTechBandit/skills/<skill-name>

### Manual install

​```bash
git clone https://github.com/BadTechBandit/skills.git
ln -s $(pwd)/skills/<skill-name> ~/.claude/skills/<skill-name>
ln -s $(pwd)/skills/<skill-name> ~/.codex/skills/<skill-name>
​```

## How it works

<File tree + 1-3 sentences on the workflow.>

## License

MIT

6. Update the main README

Two places to edit at the top of this file:

a. Add a row to the ## Skills table:

| [skill-name](./skill-name/) | <Short description, ~1-2 sentences> |

b. Add an install block under ## Install a single skill:

### skill-name

​```bash
npx skills add https://github.com/BadTechBandit/skills --skill skill-name
​```

Keep both lists in the same order. Newest skills at the top of the table = newest install block at the top of the install section.

7. Do not push

Stop after files are written and the table is updated. Show Roman a summary of what changed. He reviews and pushes.