Skill marketplace + ready-made skills
Publish + install versioned skills through a registry. Four new ready-made skills.
Skills are prompts + behavior packaged for reuse. @agentskit/skills
now ships a tiny marketplace primitive β publish semver-pinned
SkillPackages, query them, install the latest matching range β
and four new ready-made skills on top of the existing researcher /
coder / planner / critic / summarizer set.
#Install
npm install @agentskit/skills#Ready-made skills (S24 additions)
| Skill | Purpose |
|---|---|
codeReviewer | Rigorous PR review with severity-tagged findings. |
sqlGen | Natural language β parameterized Postgres queries. |
dataAnalyst | Hypothesize β query β interpret business data. |
translator | Faithful translation that preserves formatting. |
import { codeReviewer, sqlGen, dataAnalyst, translator } from '@agentskit/skills'
import { createRuntime, composeSkills } from '@agentskit/runtime'
const runtime = createRuntime({
adapter,
systemPrompt: codeReviewer.systemPrompt,
})#Marketplace primitives
import {
createSkillRegistry,
parseSemver,
compareSemver,
matchesRange,
} from '@agentskit/skills'
const registry = createSkillRegistry()
await registry.publish({
version: '1.0.0',
publisher: 'acme',
tags: ['ops'],
skill: myOpsBot,
})
// Install the latest ^1 version:
const pkg = await registry.install('ops-bot', '^1.0.0')Range syntax:
1.2.3(exact)^1.2.3(same major)~1.2.3(same minor)>=1.2.3(min version)*(any)
Enough for a basic marketplace. Layer semver on top if you need
full npm-compatible ranges. Bring your own backing store (Postgres
table of (name, version) rows, S3 manifest + CDN, Git-backed, etc.)
by implementing the same SkillRegistry contract.
#See also
- Brainstorm / compose skills
- Agentskit AI β generate a skill from natural language
Explore nearby
- PeerRecipes
Copy-paste solutions grouped by theme. Every recipe end-to-end, runs as written.
- PeerCustom adapter
Wrap any LLM API as an AgentsKit adapter. Plug-and-play with the rest of the kit in 30 lines.
- PeerAdapter contract tests
Verify any adapter against the ADR 0001 invariants A1βA10 with the shared test harness.
Browser-only chat (WebLLM / WebGPU)
Ship a chat agent that runs 100% in the user's browser β no server, no API key, no telemetry. Privacy contract holds because no inference data ever leaves the device.
Open specs β A2A, Manifest, Eval Format
Three small, versioned specs so agents, skill packs, and eval datasets travel across tools.