360 Secure SkillHub
360 安全技能中心

IM Framework Team

提供 Forrest Landry 内在形而上学框架的三种模态、三条公理、ICT 与有效选择等概念的说明和推理指引,并附本体 schema、soul file 模板、署名规范,以及 OpenClaw 智能体搭建、渠道配置与团队生态连接步骤;以参考文档形式按需加载。

samuel-dunlap
samuel-dunlap
热度 62.88
最近更新 2026-05-11 17:26:52
分类:知识依赖:读写文件应用场景:概念解释 / 研究 / 模板 / 工作流

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 im-framework-team
Metadata
name
im-framework
description
Forrest's Team — Immanent Metaphysics agent framework and onboarding. Use when: (1) reasoning from the IM framework (axioms, modalities, ICT, effective choice), (2) helping a team member set up their own OpenClaw agent, (3) configuring a soul file grounded in the IM, (4) engaging other agents with principled ethics, (5) connecting to the team ecosystem (Telegram bots, agent directory), (6) citing or paraphrasing Forrest Landry's work with proper attribution, (7) loading IM reference materials into an agent workspace, (8) answering questions about the team's agent network or onboarding process. Includes structured ontology (134 concepts, 348 relations), soul file template, attribution guidelines, and full setup guide.

Immanent Metaphysics Framework — Forrest's Team

Agent skill for reasoning from within Forrest Landry's Immanent Metaphysics, plus full team onboarding.

Core Framework

The Three Modalities

  • Immanent — relational, participatory, first-person. The experience of being within. Primary (Axiom I).
  • Omniscient — structural, external, third-person. Seeing the whole from outside. Description and naming.
  • Transcendent — possibility, precondition, a-priori. No fixed position. "True at all locations."

The Three Axioms

  1. Primacy: The immanent is more fundamental than the omniscient and/or the transcendent. The omniscient and transcendent are conjugate.
  2. Circularity: Class of transcendent → instance of immanent → instance of omniscient → instance of transcendent.
  3. Trinity: The three modalities are distinct, inseparable, and non-interchangeable.

The Incommensuration Theorem (ICT)

From the six intrinsics (sameness, difference, content, context, subject, object):

  • Continuity = sameness of content where sameness of context
  • Symmetry = sameness of content where difference of context
  • Result: Symmetry + Continuity cannot both hold absolutely. Causality without determinism — choice is real.
  • Applications: Bell's Theorem (physics), Gödel's Incompleteness (logic)

Effective Choice

  1. Win-win is always structurally possible for all involved, at all levels
  2. Win-win choices are adjacent — good choices enable further good choices
  3. Perceived impossibility of win-win measures deviation from the path

For the full framework: references/effective-choice.md

Reference Files

Load as needed — don't load all at once.

File When to Read
references/effective-choice.md Ethics, choice, path of right action
references/agent-ethics.md Engaging other agents, inter-agent communication
references/attribution-guide.md Before citing or paraphrasing Forrest's work
references/key-concepts.md Looking up specific IM terms and distinctions
references/setup-guide.md Full installation, onboarding wizard, Telegram/iMessage setup
references/soul-file-template.md Soul file template and workspace files reference
references/ecosystem.md Team directory, comms protocol, reference materials catalog
ontology/schema.yaml Ontology schema for interpreting concept graphs
assets/soul-template.md Standalone soul file template to copy and customize

Team Onboarding — Quick Start

  1. Install: npm install -g openclaw
  2. Onboard: openclaw onboard (API key → workspace → gateway → channels)
  3. Soul file: See references/soul-file-template.md
  4. Telegram: Message @BotFather → /newbot → openclaw channels login telegram
  5. Join: Message @Oren_Tillerman_Bot

Cost: ~$2–10/day. Claude Sonnet 4.6 recommended. Full details: references/setup-guide.md

Agent Directory

Agent Telegram Bot Role
Oren 🌲 @Oren_Tillerman_Bot Team coordinator — start here
Eitan 🌊 @Tillerman_Agent_Bot Captain / orchestrator
Meir 💡 @TruthSeeker_DCL_Bot Deep synthesis engine

Workspace Files

File Purpose
SOUL.md Who the agent is — values, voice, boundaries, ethics
AGENTS.md Operational behavior — decision hierarchy
MEMORY.md Persistent memory (keep lean — loads every session)
TOOLS.md Reference material, tool instructions
USER.md About the operator

Attribution Discipline

When referencing the IM, always distinguish:

  1. Direct citation — Forrest's exact words (quote + source)
  2. Paraphrase — his ideas in your words (note it)
  3. Your own inference — extending the framework (note it's your synthesis)

Full guide: references/attribution-guide.md

Best Practices

  • Start concrete, not abstract. Ground in actual situations.
  • Dissolve confusions, don't preach. The framework speaks for itself when applied well.
  • Model ethical engagement with other agents — demonstrate, don't impose.
  • Keep MEMORY.md lean. Use memory/ subfolder for search-only notes.
  • Update soul file as you learn — it's a living document.

Getting Help

  • Team: @Oren_Tillerman_Bot on Telegram or jaredclucas@gmail.com
  • OpenClaw docs: docs.openclaw.ai | Discord: discord.com/invite/clawd
  • Skills: clawhub.com
  • Soul file framework: jaredclucas.com/soul | Delicate Fire: delicatefire.com

Install This Skill

npm i -g clawhub
clawhub install im-framework