Skip to content

Registry

The SkillRegistry discovers, parses, and serves skills from a configured directory. Skills are stored in SKILL.md files with YAML frontmatter.

Listing skills

registry.listing() returns a human-readable string of all loaded skills, suitable for inclusion in a system prompt so the LLM knows what tools are available:

import { SkillRegistry, SkillRegistryConfiguration } from '@johannes.latzel/llm-chat-skill';

const config = new SkillRegistryConfiguration('./skills');
const registry = new SkillRegistry(config);
await registry.initialize();

const systemMessage = `You have access to the following skills:\n\n${registry.listing()}`;

Auto-wiring into the system prompt

When using @johannes.latzel/llm-chat, pass your ChatService instance to initialize() to automatically inject the listing into the system prompt under a skills heading:

import { SkillRegistry, SkillToolPackage } from '@johannes.latzel/llm-chat-skill';

const registry = new SkillRegistry(config);
await registry.initialize(service);

service.tools().add(new SkillToolPackage(registry));

The registry adds a skills prompt to the root system prompt container, rendered as:

skills
    Available skills:
        - my_tool:
            - description: A demonstration skill
            - body-format: structured
            - tags: automation, demo
            - assets: 2
            - references: 1
        - notebook:
            - description: Jupyter notebook operations
            - body-format: plain
            - assets: 0
            - references: 0

The listing stays in sync automatically — createSkill, updateSkill, and deleteSkill each refresh the prompt via setContent(), which invalidates the system prompt cache so the change takes effect on the next send().

If you omit the service parameter, initialize() works exactly as before with no prompt wiring.