TextSkillLoaderDefinition

KIT · interface · Type contract

Loader limits, parsing hooks and rejection wording.

Import

TypeScript

import type { TextSkillLoaderDefinition } from '@jangts/agent-core/kits/text-skills';

JavaScript

/** @typedef {import('@jangts/agent-core/kits/text-skills').TextSkillLoaderDefinition} TextSkillLoaderDefinition */

Other public entry points:@jangts/agent-core, @jangts/agent-core/types, @jangts/agent-core/structs, @jangts/agent-core/kits/text-skills, @jangts/agent-core/kits.

Definition

interface TextSkillLoaderDefinition

Field structure

Field	Type	Required	Description
maxInstructionChars	number	Required	Read-only. Maximum characters of the instruction body; larger documents are rejected.
maxDescriptionChars	number	Required	Read-only. Maximum characters of the description header.
maxIdChars	number | undefined	Optional	Read-only. Maximum id length; defaults to 64.
entryFor	((id: string) => string) | undefined	Optional	Read-only. Maps an id to its expected document path; defaults to <id>/SKILL.md.
parseHeader	((instructions: string) => TextSkillHeader | undefined) | undefined	Optional	Read-only. Default parser accepts the strict name/description header; hosts can inject a YAML parser.
tags	readonly string[] | undefined	Optional	Read-only. Diagnostic tags attached to loader rejections.
errors	{ readonly catalog?: BreakReason; readonly registration?: BreakReason; readonly document?: BreakReason; } | undefined	Optional	Read-only. See the type definition and usage context.

Examples

• TextSkills: read-only text Skills: Validate registered text work instructions that the App reads; code inside the files is never executed.

Read one text Skill from an explicit source

Beginner · Node.js

TypeScript

import { TextSkills } from '@jangts/agent-core/kits/text-skills';
const documents = new Map([['summarize/SKILL.md', '---\nname: summarize\ndescription: 汇总已读取资料\n---\n只根据已确认资料总结。']]);
const loader = TextSkills.defineTextSkillLoader({ maxInstructionChars: 2000, maxDescriptionChars: 120 });
const skills = loader.load({ entries: [{ id: 'summarize', entry: 'summarize/SKILL.md', kind: 'text-workflow', toolIds: [] }] }, {
  read(entry) { const text = documents.get(entry); if (!text) throw new Error('来源未登记'); return text; },
});
console.log(skills[0].description);

JavaScript

import { TextSkills } from '@jangts/agent-core/kits/text-skills';
const documents = new Map([['summarize/SKILL.md', '---\nname: summarize\ndescription: 汇总已读取资料\n---\n只根据已确认资料总结。']]);
const loader = TextSkills.defineTextSkillLoader({ maxInstructionChars: 2000, maxDescriptionChars: 120 });
const skills = loader.load({ entries: [{ id: 'summarize', entry: 'summarize/SKILL.md', kind: 'text-workflow', toolIds: [] }] }, {
    read(entry) { const text = documents.get(entry); if (!text)
        throw new Error('来源未登记'); return text; },
});
console.log(skills[0].description);

Add the application's own entry convention to the same loader

Configuration · Node.js

Entry rules and tool references are only registration checks; the actual tools still need to be registered and authorized by the application.

TypeScript

import { TextSkills } from '@jangts/agent-core/kits/text-skills';
const documents = new Map([['skills/summarize.md', '---\nname: summarize\ndescription: 汇总已读取资料\n---\n只根据已确认资料总结。']]);
const loader = TextSkills.defineTextSkillLoader({ maxInstructionChars: 2000, maxDescriptionChars: 120, maxIdChars: 32, entryFor: id => 'skills/' + id + '.md' });
const skills = loader.load({ entries: [{ id: 'summarize', entry: 'skills/summarize.md', kind: 'text-workflow', toolIds: ['read-source'] }] }, {
  read(entry) { const text = documents.get(entry); if (!text) throw new Error('来源未登记'); return text; },
});
console.log(skills[0].toolIds);

JavaScript

import { TextSkills } from '@jangts/agent-core/kits/text-skills';
const documents = new Map([['skills/summarize.md', '---\nname: summarize\ndescription: 汇总已读取资料\n---\n只根据已确认资料总结。']]);
const loader = TextSkills.defineTextSkillLoader({ maxInstructionChars: 2000, maxDescriptionChars: 120, maxIdChars: 32, entryFor: id => 'skills/' + id + '.md' });
const skills = loader.load({ entries: [{ id: 'summarize', entry: 'skills/summarize.md', kind: 'text-workflow', toolIds: ['read-source'] }] }, {
    read(entry) { const text = documents.get(entry); if (!text)
        throw new Error('来源未登记'); return text; },
});
console.log(skills[0].toolIds);

Related policy source

• Text registration and strict header parsing

Related definitions

• BreakReason
• TextSkillHeader