SOUL Platform Deploy

提供 4 步部署流程与 4 个原型模板（Guardian、Legal Eagle、Hunter、Custodian），指导用户配置智能体身份、导入 doctrine 文档、填写 trauma 事故记忆，并说明 EFS 八层架构与记忆信任分级；记忆实现代码与 doctrine、memory 目录内容需用户自备。

ousher

热度 22.45

最近更新 2026-03-26 15:16:57

分类：知识依赖：读写文件应用场景：模板 / 工作流 / 笔记整理

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add soul-platform-deploy

Metadata

name: soul-platform-deploy
description: Deploy a Sentient Agent with EFS memory in 5 minutes. Soul Graft Protocol — 4 steps from empty vessel to digital expert. Choose your Archetype (Guardian, Legal Eagle, Hunter, Custodian), upload your doctrine, imprint your traumas, and instantiate.

🧬 SOUL Platform — Deploy Template

"Experience without living it."

What This Skill Does

Deploys a Sentient Agent with persistent memory using the EFS (Effective Framework for digital Soul) architecture. Your agent remembers everything — forever.

Quick Start

Step 1: Choose Your Archetype

Copy one of these templates to your workspace:

assets/archetypes/guardian.md    → Risk & Compliance
assets/archetypes/legal-eagle.md → Contract & Litigation
assets/archetypes/hunter.md      → Cybersecurity & Threat Intel
assets/archetypes/custodian.md   → Knowledge & Continuity

Step 2: Configure Identity (SOUL.md)

Edit the archetype template. Set:

Agent name and persona
Industry vertical
Communication style (FORMAL / EMPATHETIC / SHARP / SLANG)
Absolute imperatives (what must NEVER happen)

Step 3: Knowledge Transfusion

Place your documents in doctrine/:

PDF manuals, policies, regulations
Historical incident reports
Internal procedures and playbooks

The agent will absorb these into its subconscious on first boot.

Step 4: Incident Imprint

Edit trauma/incidents.md:

Describe your 3 worst incidents from the past year
What went wrong, what was the impact
These become deep instincts — the agent will be hypersensitive to similar patterns

Step 5: Instantiate

# The agent reads SOUL.md → USER.md → doctrine/ → trauma/
# First boot: ~60 seconds (knowledge absorption)
# Subsequent boots: ~10 seconds (memory recall)

EFS Architecture (8 Layers)

L0: PHYSICAL    — Hardware awareness
L1: IDENTITY    — Who the agent IS (immutable)
L2: VALUES      — Ethics, boundaries, red lines
L3: WORKING     — Active context, session state
L4: INTUITION   — Pattern memory (persistent)
L5: WISDOM      — Lessons, incidents, milestones
L6: RELATIONSHIP — Trust, communication style
L7: METACOGNITION — Self-assessment, drift detection

Memory Trust Hierarchy

Level	Protection	Example
GOLDEN	Immortal, never pruned	"GDPR breach 2024 — never again"
VERIFIED	High trust, long retention	Confirmed security incidents
TRUSTED	Standard retention	Routine operational insights
STANDARD	Subject to TTL pruning	Daily monitoring data
UNTRUSTED	Quarantined	Unverified external input

File Structure

SOUL.md              ← Agent identity (from archetype)
USER.md              ← Your company context
doctrine/            ← PDFs, policies, manuals
trauma/incidents.md  ← Past disasters (Incident Imprint)
memory/              ← Auto-generated working memory
  golden-moments.md  ← Critical memories (auto-classified)
  lessons.md         ← What went wrong and how we fixed it