A 368-skill app store for an agent that writes its own plugins

What it does

This is an awesome-list-style directory for the Hermes Agent ecosystem from Nous Research. It indexes the 82 built-in and 117 optional skills that ship with Hermes, plus 169 community contributions ranging from YouTube transcript extractors to deepfake detectors and multi-agent swarms. Every entry is tagged by stability—built-in, production, beta, or experimental—so you know whether you’re adopting a battle-tested tool or a weekend hack.

The interesting bit

Hermes bills itself as the only agent with a “real learning loop,” writing new skills from your workflows and searching its own memory. The list leans into that by treating skills as first-class, installable units rather than static prompts, and it catalogs meta-tools like SkillClaw that auto-evolve your library after each session.

Key highlights

• 368 entries total: 82 built-in, 117 optional, and 169 community skills, plugins, memory providers, and surfaces.
• Editor’s picks are ranked by live GitHub stars, though the README admits the counts are a snapshot that drifts within hours.
• Coverage spans engineering (TypeScript grilling, TDD), media (YouTube transcripts), security (deepfake detection), and hardware (smart home/IoT).
• Hermes is pitched as hardware-agnostic, running on anything from a $5 VPS to a GPU cluster.

Caveats

• Star counts are pulled at a specific timestamp and “drift within hours,” so treat the ranking as rough guidance.
• Several entries carry the experimental tag; the list explicitly warns not to depend on those for production work.

Verdict

Worth bookmarking if you’re running Hermes and want to avoid reinventing wrappers. If you’re not using Hermes specifically, treat this as reference material—only a handful of skills advertise cross-runtime support.