5 free GitHub skills that replaced my entire content team
Blog posts, images, ghostwriting, email sequences, YouTube strategy. Each one a slash command. Each one free. Here's how to actually implement them — including the step most people skip.
I had a client who used to spend around £1,400 a month on content. A blog writer, a ghostwriter for LinkedIn, a designer for visuals, a strategist for YouTube. Not because the output was great — it was fine. Because they didn't have a better option.
Then I found these five GitHub skills for Claude Code and implemented an entire system for them. They're not prompt packs. They're structured methodology documents that load into Claude's context and make it operate like a specialist. The difference in output quality is immediate and obvious.
Here's how to install and — more importantly — how to use each one correctly. The implementation step most people skip is the one that makes the skill actually work.
Most AI blog tools output text and leave you to guess whether it's any good. claude-blog outputs a 100-point quality report alongside the post — scoring Content Quality, SEO, E-E-A-T signals, Technical SEO, and AI Citation Readiness separately.
The citation readiness score is the differentiator. It measures whether your post is likely to be quoted by ChatGPT, Perplexity, or Google AI Overviews. Most content skips this entirely.
You guess at SEO
No quality signal
AI phrases not checked
~10 min to format
No follow-up commands
Title/meta validated
84/100 scored, auto
17-phrase scan, 0 flagged
47 seconds end-to-end
/blog repurpose built in
/blog analyze [competitor-url] on the top-ranking post for your target keyword. The skill reads their structure — heading hierarchy, stat density, FAQ pattern, word count — and uses it as the quality baseline for your post. Skip this and you're writing into a vacuum. Do it and you're targeting a known standard.$ gh repo clone AgriciDaniel/claude-blog ~/.claude/skills/claude-blog
# Analyse competition first (the step most skip)
$ /blog analyze https://competitor.com/their-top-post
Reading structure... heading depth: 4, stat count: 11, FAQ: yes
✓ Baseline saved. Will target this quality standard.
$ /blog write "your topic here"
✓ Quality score: 84/100 · 9 sources · 0 AI phrases · 47s
The problem with AI image generation isn't the model — it's that most people pass their rough idea directly to the API and wonder why the output is mediocre. Banana fixes this by acting as a Creative Director that converts your brief into a 300-word structured prompt before touching Gemini.
It runs inside Claude Code. No new app, no Midjourney subscription, no extra tab. You type 30 words. It writes 300. Gemini generates the image. Total cost: ~$0.13.
$ /banana generate
Describe your image (or describe someone describing it):
> creative director briefing: solo founder, white t-shirt, 5 glowing
terminal windows fanned like cards. Text: "5 Skills. No Team."
Building 300-word creative brief...
✓ Generated · $0.13 · Text rendered correctly
18 stars. The most underrated skill on this list. It's been doing the work of three content freelancers (£400/month each, combined £14,400/year) for the cost of API tokens.
The skill runs a structured voice interview before writing anything. It asks how you write, what you avoid, your banned phrases, your sentence length preference. It builds a voice fingerprint. Then every post it writes is run through that fingerprint before it leaves Claude's output.
# Calibrate with YOUR real writing (not AI-edited posts)
$ /ghost calibrate
Paste 3–5 posts you wrote. I'll extract your voice fingerprint.
✓ Voice fingerprint saved · Banned phrases: 0 found in your writing
$ /ghost write linkedin "my take on [topic]"
✓ Voice match: HIGH · 0 banned phrases · British English
Most email sequence generators produce generic copy because they don't know anything about your audience. gtm-agents is different because it loads 244 structured methodology files — including Value Ladder frameworks, signal-branching cadences, and ICP modelling — before writing a single word.
The output isn't a prompt. It's a 3-email sequence with send timing built in, objection handling baked into Email 2, and a DM trigger mechanism in Email 3. Under 2 minutes from command to ready-to-send.
CONTEXT.md file in your project directory before running any sequence. Include: your ICP (job title, company size, pain point), your lead magnet title, your main competitor, and 2–3 objections you hear on sales calls. The skill reads this file automatically and bakes it into every email. Without it: solid generic copy. With it: output specific enough that prospects assume you know them personally.Role: Founder, 5–50 person SaaS
Pain: spending £1k+/month on content with no ROI signal
# Lead Magnet
Title: "5 Claude Code Skills That Replaced My Content Team"
# Top Objections
- "AI content sounds generic"
- "I don't have time to set this up"
Reading CONTEXT.md... ICP locked. Objections loaded.
✓ Email 1 (immediate) — deliver lead magnet + social proof
✓ Email 2 (+48h) — handle "AI sounds generic" objection
✓ Email 3 (+72h) — DM trigger: reply SKILLS
Generated in 94 seconds.
YouTube strategy tools usually do one thing. claude-youtube runs a full content strategy in one command — 5 hook variants with psychological trigger labels, a thumbnail brief with exact hex codes and mobile preview, a 3-act script outline with retention risk table, and an SEO panel with 15 tags and 3 title variants.
The skill knows which hook mechanism works for which traffic source (search vs suggested vs external) and surfaces a recommendation based on your channel's data.
/youtube strategy on a new video, run /youtube analyze on your 3 best-performing videos first. The skill reverse-engineers what worked — which hook type, which thumbnail composition, which title pattern — and applies those patterns as priors to your new content. Without this step, the skill treats all channels identically. With it, the recommendations are calibrated to what your specific audience actually responds to.$ /youtube analyze https://youtu.be/your-best-video
✓ Hook type: Story-Open · Thumbnail: face + text · CTR: 8.2%
Pattern saved. Will bias new strategy toward Story hooks.
$ /youtube strategy "5 Claude Code skills that replaced my content team"
✓ 5 hooks · Recommended: Story-Open (matches your top performer)
✓ Thumbnail: hex codes + mobile preview + safe zone check
✓ 15 tags · 3 title variants · Recommended title highlighted
Run them together and you have a content system
Individually each skill is useful. Stacked in order, they become a content production pipeline. This is the exact workflow I run for every piece of content now:
That's a blog post, a LinkedIn post, a YouTube strategy, a 3-email sequence, and a custom visual. Six commands. No agency. No subscriptions. API costs for the whole run: under £2.
Install all five
$ gh repo clone AgriciDaniel/claude-blog ~/.claude/skills/claude-blog
# 2. banana — AI image creative director (MCP)
$ npm install -g @ycse/nanobanana-mcp
# 3. founder-voice-ghostwriter — writes like you
$ gh repo clone BayramAnnakov/founder-voice-ghostwriter ~/.claude/skills/ghostwriter
# 4. gtm-agents — email sequences + GTM
$ gh repo clone gtmagents/gtm-agents ~/.claude/skills/gtm-agents
# 5. claude-youtube — hook variants + strategy
$ gh repo clone AgriciDaniel/claude-youtube ~/.claude/skills/claude-youtube
✓ All 5 installed. Restart Claude Code to load.