Hermes Agent Skills Hub: Find and Install Skills

The Hermes Agent skills hub is the registry where you browse, search, inspect, and install reusable capabilities for the agent, built on the agentskills.io open standard. In Hermes Agent, you reach it from the command line with hermes skills browse or in chat by typing /skills, then install any skill with a single command like hermes skills install official/migration/openclaw-migration.

What the Hermes Agent skills hub is

The Hermes Agent skills hub is a registry of portable, shareable, community-contributed skills that the agent can discover and install on demand. Hermes Agent is an open-source, self-hostable AI agent built by Nous Research and released in February 2026 under the MIT license.

The hub is reachable two ways. From the command line you run hermes skills browse to page through skill registries; inside a chat session on Telegram, Discord, Slack, or the CLI you type /skills to list what is available. Both surfaces draw on the same agentskills.io open standard, so skills you find are not locked to a single vendor.

Hermes also creates skills on its own. According to the official documentation, after solving a hard problem the agent writes a reusable skill document so it never has to relearn the same task — what Nous calls procedural memory.

What skills are and the SKILL.md format

A skill is a directory containing, at minimum, a SKILL.md file with YAML frontmatter followed by Markdown instructions. The format is defined by the agentskills.io specification, an open standard supported by 27+ AI agents including Claude Code, Codex, Cursor, and Gemini CLI.

The frontmatter requires only two fields. The name field is 1–64 lowercase alphanumeric characters and hyphens and must match the parent directory name. The description field (up to 1024 characters) explains what the skill does and when to use it. Optional fields include license, compatibility, metadata, and the experimental allowed-tools.

Skills load through progressive disclosure. Per the specification, the name and description (~100 tokens) are loaded at startup for every installed skill; the full SKILL.md body (recommended under 5,000 tokens) loads only when a task matches; and files in scripts/, references/, or assets/ load only when actually required. This keeps the agent's context lean even with many skills installed.

How to browse, search, and inspect skills

Hermes Agent exposes four discovery commands before you commit to installing anything. They let you page through registries, search by keyword or documentation URL, and preview a skill's full details first.

The inspect step matters. Because skills can ship executable scripts, reading the body and checking the source before installing is the cheapest safety check available. You can inspect a registry entry like official/security/1password or a raw URL such as https://sharethis.chat/SKILL.md.

How to install and manage skills

You install a Hermes skill with hermes skills install, pointing it at a registry path or a direct URL. The agent also detects bundled skills automatically and seeds them on setup.

# Install from the official registry
hermes skills install official/migration/openclaw-migration

# Install from a direct URL with a custom name
hermes skills install https://example.com/SKILL.md --name my-skill

# Install a third-party skill, overriding policy blocks
hermes skills install skills-sh/anthropics/skills/pdf --force

Once installed, skills live at ~/.hermes/skills/, and skills imported from OpenClaw land in ~/.hermes/skills/openclaw-imports/. Day-to-day management uses a small set of commands: hermes skills list shows what is installed, hermes skills check looks for upstream updates, hermes skills update reinstalls hub skills with the latest changes, and hermes skills config enables or disables skills per platform. To undo local edits, hermes skills reset <name> clears your modifications and --restore reverts to the bundled version.

If you want to write your own skill rather than install one, see our companion Hermes Agent skills guide, which walks through creating and managing custom skills end to end. The agentskills.io skills-ref validate ./my-skill tool checks your frontmatter against the spec before you publish.

How Remote OpenClaw's /skills directory helps

Remote OpenClaw's skills directory is a browsable catalog that helps you discover OpenClaw-lineage skills before you install them into Hermes Agent or any agentskills.io-compatible agent. Because Hermes can install from a direct SKILL.md URL, a catalog that surfaces well-described skills shortens the search step the CLI alone leaves to you.

The directory groups skills so you can scan by purpose instead of guessing registry paths — useful when you know the outcome you want but not the exact skill name. Once you find a candidate, you preview it, then install it with hermes skills install or by inspecting its raw URL. For a wider comparison of where these two ecosystems overlap, our OpenClaw vs Hermes Agent breakdown covers memory, skills, and best fit.

Remote OpenClaw is a content hub, marketplace, and community for AI agents — it sells personas, skills, and bundles and curates discovery surfaces like /skills. It does not host or deploy your agent; you self-host Hermes yourself following the setup guide.

Limitations and Tradeoffs

Skills are powerful but carry real tradeoffs. The biggest is trust: skills can bundle executable scripts/, so installing one from an unverified URL is equivalent to running someone else's code. Always hermes skills inspect a skill and read its source before installing, and treat --force (which overrides policy blocks) as a deliberate choice, not a default.

Do not over-install. Every skill's metadata loads into context at startup, so a sprawling library adds noise even with progressive disclosure. Install what you use, disable the rest with hermes skills config, and remove unmodified bundled skills with opt-out --remove if you want a leaner setup.

Finally, the standard is young. The allowed-tools field is explicitly experimental and support varies between agents, and command syntax can change between releases — these commands reflect Hermes v0.17.0 (June 19, 2026). When in doubt, check the official CLI reference for your version.

Related Guides

• Hermes Agent Skills: How to Create and Manage Custom Skills
• How to Install Hermes Agent
• Hermes Agent MCP Integration: Using Model Context Protocol
• OpenClaw vs Hermes Agent: Memory, Skills, and Best Fit