Complete cognitive architecture for the Hermes Agent ecosystem — a self-evolving, spec-driven, multi-agent knowledge management system.
Seven Layers
This architecture defines seven interconnected layers that together form a persistent, autonomous agent identity with durable memory, procedural skills, and spec-driven execution capabilities.
| Layer | Section | Description |
|---|---|---|
| 1 | Identity Anchors | Core identity, ethics, persona, social contracts |
| 2 | Mutable Memory | Semantic, episodic, session, and feedback memory |
| 3 | Orchestration & Rules | Agent coordination, conventions, safety gates |
| 4 | Capabilities & Infrastructure | Skills, procedures, tools, checklists |
| 5 | Autonomic Triggers | Cron, bootstrap, boot hooks |
| 6 | Spec-Driven Working Memory | Plans, briefs, requirements, tasks |
| 7 | Swarm & Domain Alignment | Design system, thesis, signals |
Quick Reference
| File | Purpose | Location |
|---|---|---|
SOUL | Ethical constraints, escalation paths | identity-anchors/soul.md |
USER | Human operator profile | identity-anchors/user.md |
IDENTITY | Bot naming, avatars, routing | identity-anchors/identity.md |
MEMORY | Curated long-term facts | mutable-memory/memory.md |
AGENTS | Multi-agent coordination | orchestration/agents.md |
SKILL | Procedural skill manuals | capabilities/skill.md |
HEARTBEAT | Cron schedules, health checks | autonomic/heartbeat.md |
PLAN | Active goal tracking | spec-driven/plan.md |
THESIS | Strategic north star | swarm-alignment/thesis.md |
This architecture is self-describing — the wiki is the system it documents. Changes to these pages propagate into agent behavior via the research automation pipeline.