@NousResearch: Hermes Agent launches the /learn instruction for automatic skill extraction. Core features and the /learn instruction launched by Nous Research, Hermes Ag…
🌐 This transcript was automatically translated to English from the original.
Hermes Agent has launched a new slash command called /learn, designed to automatically extract and build skills. The Hermes Agent, released by Nous Research, simplifies the skill creation process through this new feature. Users no longer need to manually write Skill.md files; they only need to provide raw data such as directories, code, API documentation, PDF files, or user manuals. The agent will then automatically collect and analyze this data and generate verifiable skills in a standard format. This feature supports multiple methods, such as providing local paths—for example, specifying a project folder—or directly entering URLs of online documents. It can also convert recent conversation context into skills, such as by simply stating the process of just deploying a test server. Users can even paste notes or text descriptions directly. In terms of technical operation, this system demonstrates how to process data and perform validation. The /learn command does not operate via a separate extraction engine; instead, it is handled directly by the agent using existing tools like Read File or Web Extract. This ensures a consistent user experience across different interfaces such as CLI, TUI, or Dashboard. The generated skills adhere to standardized specifications, such as a description length limit of 60 characters and a standard chapter ordering, before being stored via the Skill Underscore Manage tool. If the system has collaborative review enabled, the skills must be confirmed by the user before they are officially stored. The Hermes Agent's skill system is designed as on-demand retrieval, which effectively reduces token usage and is compatible with the open standards of AgentSkills.io. By default, all skills are stored in the .hermes-skill folder under the user's home directory. If users wish to customize their environment, they can manage default skill combinations via commands—for example, excluding default skills during installation or selecting exclusions when creating a configuration file. Adjustments can also be made during execution, such as using an Opt-out command to stop future automatic synchronization or adding a Remove parameter to delete unmodified default skills. Of course, an Opt-in command can also be used to re-enable and resynchronize. This system ensures high extensibility for the agent, allowing users to quickly accumulate personalized automated workflows through the /learn command and call upon other commands starting with a slash, such as /plan or /excalibur, at any time.