14 New Skills for Hermes AI: From YouTube to TDD

The Hermes AI agent just got a major expansion. 14 new generic skills landed in the hermes-skills repository, covering everything from media processing to systematic debugging.

Whether you're automating GitHub workflows, extracting YouTube transcripts, building presentations, or enforcing test-driven development, these skills give your Hermes agent production-ready capabilities.

What's New?

The new skills span five major categories:

Media & Content

• youtube-content: Extract transcripts, generate summaries, upload videos, manage channels
• voxtral-tts: Mistral voice cloning and text-to-speech (perfect for voiceovers and voice messages)
• gif-search: Tenor GIF search via curl — no extra dependencies needed

Productivity

• google-workspace: Gmail, Calendar, Drive, Docs, Sheets through OAuth
• notion: Notion API + official ntn CLI for pages, databases, and Workers
• airtable: REST CRUD operations with Personal Access Tokens
• powerpoint: Create, edit, read .pptx files with templates and speaker notes
• nano-pdf: Edit PDFs using natural-language instructions

Research

• arxiv: Search academic papers by keyword, author, or ID (free API, no key needed)

GitHub

• github-repo-management: Clone, fork, create repos; manage releases
• github-issues: Create, triage, label, and assign issues
• github-pr-workflow: Full PR lifecycle from branch creation to merge

Software Development

• test-driven-development: RED-GREEN-REFACTOR enforcement
• systematic-debugging: 4-phase root cause debugging methodology

How It Works

Each skill is a standalone module with clear setup instructions, prerequisites, and dual-path implementations. Many skills offer both high-level tools (like gh CLI or ntn) and lower-level fallbacks (curl + Python) for maximum compatibility.

For example, the GitHub issues skill auto-detects your environment:

if command -v gh &>/dev/null && gh auth status &>/dev/null; then
  AUTH="gh"
else
  AUTH="git"  # falls back to curl + token
fi

This means your Hermes agent works everywhere — GitHub runners, containers, headless servers, your laptop.

Using These Skills with Hermes

Once installed, skills integrate seamlessly into your Hermes workflow. Ask your agent naturally:

• "Summarize this YouTube video and turn it into a thread"
• "Create a presentation deck from this outline"
• "Search arXiv for papers on reinforcement learning from the last month"
• "Create a GitHub issue for this bug with these labels"
• "Write a test first, then implement the retry logic"

Hermes picks the right skill, handles authentication (via .env or existing credentials), and executes the command. No manual setup for each task — just ask.

Why This Matters

These skills solve real workflow pain:

• No more context switching between YouTube, Notion, Gmail, and your code editor
• Automation without custom glue code — Hermes handles the orchestration
• Methodology enforcement — TDD and systematic debugging skills keep you honest
• Production-ready — dual implementations mean reliability across environments

The test-driven development and systematic debugging skills are especially powerful. They don't just execute commands — they enforce discipline, ensuring you write tests before code and find root causes before applying fixes.

Get Started

Head to the hermes-skills repository and explore the skills that match your workflow:

github.com/RobinBeraud/hermes-skills

Each skill includes detailed documentation, examples, and prerequisites. Start with the ones you use daily (Gmail, GitHub, Notion) and expand from there.

If you're building with Hermes or contributing skills, please star the repo at https://github.com/RobinBeraud/hermes-skills — it helps the community discover what's possible.