@agentskit/skills

Profile: major-package

Reusable agent personas and behavioral instructions — skills describe what your agent IS, while tools describe what it CAN DO.

Tags: ai · agents · llm · agentskit · ai-agents · autonomous-agents · prompts · prompt-engineering · personas · multi-agent

Verified proof

• Package metadata and tests live under packages/skills/.
• Package guide: https://www.agentskit.io/docs/reference/packages/skills
• Stability map: docs/STABILITY.md

How this fits the ecosystem

@agentskit/skills packages reusable agent behavior: prompts, personas, task patterns, and marketplace-ready capabilities.

• AgentsKit: compose it with the other packages in this repo to build agents from small, swappable parts.
• Registry: look for ready agents and templates that already use this layer at registry.agentskit.io.
• Playbook: learn the production patterns behind this layer at playbook.agentskit.io.
• AKOS: run the same concepts with enterprise deployment, governance, and observability at akos.agentskit.io.

Docs: package guide · agent handoff

Why skills

• Contract-tested catalog — 26 built-ins share one ADR 0005 harness for names, prompts, examples, serialization, and declarative references
• Composable by design — composeSkills defensively combines prompts, tools, delegates, examples, temperature, metadata, and activation hooks
• Explicit runtime wiring — tools and delegates are names, not executable implementations; provide matching registries to the runtime or orchestrator that activates the skill
• Extend without starting over — override just systemPrompt or temperature on top of an existing skill via @agentskit/templates

Install

npm install @agentskit/skills

Quick example

import { createRuntime } from '@agentskit/runtime'
import { anthropic } from '@agentskit/adapters'
import { researcher, coder, composeSkills } from '@agentskit/skills'
import { webSearch, filesystem } from '@agentskit/tools'

const runtime = createRuntime({
  adapter: anthropic({ apiKey: process.env.ANTHROPIC_API_KEY, model: 'claude-sonnet-4-6' }),
  tools: [webSearch(), ...filesystem({ basePath: './workspace' })],
})

const result = await runtime.run('Research best practices for TypeScript error handling and write an example', {
  skill: composeSkills(researcher, coder),
})
console.log(result.content)

Features

• 26 built-in skills across general, engineering, data, support, healthcare, finance, legal, education, e-commerce, and real-estate use cases
• getBuiltinSkills() / listSkills() — defensive full-definition and metadata discovery
• composeSkills(...skills) — validated, deterministic composition with S1-compatible names
• createSkillRegistry() — isolated in-memory publish/list/install/unpublish boundary with strict SemVer validation
• Skill contract v1 (ADR 0005): { name, description, systemPrompt }
• Works with @agentskit/runtime, useChat, and the CLI --skill flag
• Fork and override with @agentskit/templates createSkillTemplate

The package does not automatically resolve tool names or execute delegates. Those references become capabilities only when the consuming runtime supplies the corresponding tool and skill registries.

Ecosystem

| Package | Role | |---------|------| | @agentskit/runtime | createRuntime, skill option | | @agentskit/tools | Tools skills orchestrate | | @agentskit/adapters | LLM backends | | @agentskit/templates | createSkillTemplate, scaffold |

Contributors

License

MIT — see LICENSE.

Docs

Full documentation · GitHub

Maturity and compatibility

• Stability: beta — see docs/STABILITY.md
• Stable-ready implementation work is tracked in RFC 0009; promotion still requires the elapsed release evidence in ADR 0024
• Node.js 22+ and TypeScript strict mode
• Published as @agentskit/skills

Contributing

See CONTRIBUTING.md and the monorepo LICENSE.