Web Search
Real-time web search for Hermes agents via Tavily or Serper
hermes skills search web-searchPersistent Memory
Long-term memory store for Hermes agents across sessions
hermes skills search memoryCode Executor
Run Python and shell code safely inside Hermes agent tasks
hermes skills search code-execEmail Skill
Send, read and draft emails from your Hermes agent via Gmail or SMTP
hermes skills search emailBrowser Automation
Control a headless browser to scrape, fill forms and navigate the web
hermes skills search browserSlack Integration
Send messages, read channels and manage Slack from Hermes agents
hermes skills search slackDatabase Query
Query PostgreSQL, MySQL or SQLite databases directly from Hermes
hermes skills search databasePDF Reader & Writer
Extract text from PDFs and generate new PDF documents in Hermes
hermes skills search pdfCalendar Manager
Read and create Google Calendar events from Hermes agent workflows
hermes skills search calendarGitHub Assistant
Create issues, review pull requests and manage GitHub workflows
hermes skills search githubNotion Workspace
Create pages, update databases and search Notion from Hermes
hermes skills search notionWeb Scraper
Extract structured data from websites with anti-bot protection and proxy support
hermes skills search scraperFirewall Manager
Manage UFW and iptables rules from Hermes with threat detection
hermes skills search firewallAbout Hermes Agent
Hermes Agent is an open-source agent framework from Nous Research that runs in your terminal, a desktop app, messaging platforms and IDEs. Its defining feature is a closed learning loop: it can identify a capability it lacks, write a skill for it, test it, and keep it. From v0.12 an autonomous Curator periodically scores, merges and prunes the agent-created skill library.
Maintained by Nous Research.
How skills work
Hermes skills are on-demand knowledge documents stored in hermes/skills/, which is the single source of truth. Bundled skills are copied there on a fresh install, and hub-installed and agent-created skills land there too. Every installed skill automatically becomes a slash command, so a skill named plan is invoked with /plan.
Security model
Hermes checks third-party and community skills against policy before installing, and --force will not override a verdict it considers dangerous. Because Hermes can also write its own skills at runtime, review what the Curator has kept periodically with hermes skills list.