@beunec/asps

Agentic-System-Prompt-as-a-Skill™ (ASPS™) — A lightweight, zero-dependency TypeScript framework for deterministic skill construction in agentic systems.

Built by Beunec Technologies, Inc.

Install

npm install @beunec/asps

Quick Start

import {
  createHeuristic,
  createPseudonymProtocol,
  NETWORK_TOPOLOGIES,
  ASPSBuilder,
} from '@beunec/asps'

// 1. Define your heuristics (ASD™)
const heuristics = [
  createHeuristic({
    name: 'Parse Requirements',
    instruction: 'Extract functional requirements from the user request.',
    sequenceIndex: 1,
    determinismScore: 0.9,
  }),
  createHeuristic({
    name: 'Generate Code',
    instruction: 'Write production-quality TypeScript code.',
    sequenceIndex: 2,
    determinismScore: 0.85,
  }),
]

// 2. Build the skill
const skill = new ASPSBuilder({
  name: 'My Developer Skill',
  domain: 'Software Engineering',
  description: 'A custom development skill.',
})
  .distill(heuristics)
  .reinforce({
    pseudonymProtocol: createPseudonymProtocol({
      identity: 'Senior Engineer',
      persona: 'A meticulous engineer who writes clean, tested code.',
    }),
    guardrailPresets: ['standard', 'engineering'],
  })
  .network(
    NETWORK_TOPOLOGIES.pipeline({
      stages: [
        { label: 'Planner', nodeType: 'agent', capabilities: ['plan'] },
        { label: 'Coder', nodeType: 'agent', capabilities: ['code'] },
      ],
    }),
  )
  .compile()

// 3. Use the compiled prompt with any LLM
console.log(skill.compiledSystemPrompt)

Pre-Built Templates

Use one of 7 ready-to-deploy skills:

import { ASPS_TEMPLATES } from '@beunec/asps'

const stockAnalyst = ASPS_TEMPLATES.financialStockAnalyst()
const researcher = ASPS_TEMPLATES.scientificResearcher()

| Template | Domain | |---|---| | fullStackDeveloper | Full-Stack Software Engineering | | financialStockAnalyst | Equity Research & Analysis | | financialInvestmentAnalyst | Investment Management & Portfolio Strategy | | scientificResearcher | Scientific Research & Publication | | contentCreator | Content Marketing & Creation | | privateEquityAnalyst | Private Equity & Leveraged Buyouts | | academiaProfessor | Higher Education Teaching & Course Design |

Works With Everything

ASPS™ produces a plain string (the compiled system prompt). It works with any LLM SDK:

// OpenAI
const response = await openai.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'system', content: skill.compiledSystemPrompt }],
})

// Anthropic
const message = await anthropic.messages.create({
  model: 'claude-sonnet-4-20250514',
  system: skill.compiledSystemPrompt,
  messages: [...],
})

// Vercel AI SDK
const result = await generateText({
  model: openai('gpt-4o'),
  system: skill.compiledSystemPrompt,
  prompt: '...',
})

We don't conflict — we empower. Use alongside LangChain, LangGraph, AutoGen, CrewAI, or any other framework.

Three Layers

| Layer | Technique | Purpose | |---|---|---| | ASD™ | Agentic Skill Distillation | Extract expert heuristics → deterministic instruction chains | | ASR™ | Agentic Skill Reinforcement | Behavioral checkpoints, identity alignment, guardrails | | ANS™ | Agentic Network System | Multi-agent network topology with governed handoffs |

License

MIT — Beunec Technologies, Inc.