Hermes Advanced Memory Setup

A Hermes Agent skill that enforces structured workspace memory. Replaces scattered files and chaotic MEMORY.md entries with an organized workspace -- AGENTS.md index, project folders with README.md, and SOUL.md for persistent principles.

Quick Install

# Via Hermes CLI
hermes skill install salt-vrn/hermes-advanced-memory

# Manual
cd ~/.hermes/skills/
git clone https://github.com/salt-vrn/hermes-advanced-memory.git

What It Creates

~/.hermes/
├── SOUL.md              ← identity + rules (including workspace rules)
├── workspace/
│   ├── AGENTS.md        ← index  --  loaded every session, TOC for all memory
│   ├── project-alpha/
│   │   └── README.md
│   ├── project-beta/
│   │   └── README.md
│   └── ...
└── ...

Key Rules Enforced

1. Workspace is the only working folder. All agent files go into workspace/.
2. AGENTS.md is the index. Loaded at every session start. Without an index entry, a file is invisible to the agent.
3. Small topics → AGENTS.md entry. Major projects → separate folder + README.md.
4. SOUL.md is the source of truth. Workspace rules, key principles, and identity live there -- not just in the skill file.

Usage

For a New Agent

1. Agent reads the skill
2. Asks user to confirm profile (never guesses)
3. Creates workspace/ in profile root
4. Creates AGENTS.md -- empty index
5. Moves working files into workspace/
6. Opens existing SOUL.md and adds key principles
7. Describes the new structure to the user

For an Existing Agent

If memory is not organized, apply the skill, move scattered files into workspace/, and update the index.

Why This Matters

Without a memory system, Hermes agents store data chaotically -- files scattered across ~/.hermes/, MEMORY.md filled with garbage, crucial facts forgotten across sessions. This skill solves that by enforcing a structure that:

• Prevents context loss -- AGENTS.md is the single source of truth, loaded every session
• Scales with projects -- new project = new folder + README, never a cluttered flat directory
• Survives restarts -- SOUL.md preserves key principles across sessions and model changes
• Self-documents -- the structure itself tells the agent where everything lives

Production Notes

• NeiroHost uses this in production AI-agent hosting alongside hermes-session-maintenance and openclaw-to-hermes-migration
• The skill is bilingual (English + Russian README)
• MIT licensed, 8 commits as of June 2026

---

← Skills Catalog | Session Maintenance Setup → Powered by CorpusIQ

Integration with CorpusIQ

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects