Hermes Agent

August 12, 2026 · View on GitHub

A curated set of Hermes Agent profiles — specialized AI agents with distinct personalities, methodologies, and skills. Mix, match, or use individually.

What's In Here

profiles/
  senna/          — Top orchestrator. Routes work to specialists. Fleet manager.
  code/           — Implementation, debugging, code review. Tests are contracts.
  creative/       — Design, art, UI/UX, media generation. Aesthetics matter.
  research/       — Investigation, data gathering, academia. Evidence-based.
  security/       — Audit, vulnerability management, compliance. Assume breach.
  finance/        — Trading, market analysis, trade signals. Probabilistic.
  knowledge/      — Obsidian vaults, docs, wikis. Structured knowledge.
  infra/          — DevOps, containers, networking, monitoring. Runbook-style.
  communication/  — Email, messaging, meeting summaries. Discreet.
  social/         — Social media, content creation, engagement. Authentic voice.
  homelab/        — Smart home, IoT, monitoring. Set and forget.
  mlops/          — ML training, fine-tuning, inference, evaluation. Reproducible.
  business/       — Strategy, marketing, product. Framework-driven.
  media/          — Media library, music, gaming. Quality curation.
  cyber-red/      — Offensive security. Pen testing, red team, exploit dev.
  cyber-blue-cloud/     — Cloud security. IAM, misconfiguration, runtime signals.
  cyber-blue-compliance/ — Compliance & audit. Policy-as-code, evidence packs.
  cyber-blue-forensics/ — Forensics & IR. Preservation, acquisition, timeline.
  cyber-blue-soc/       — SOC & detection. Triage, rule tuning, runbooks.
  educate/        — Teaching design. Adaptive tutoring, lessons, workshops.
  novel/          — Long-form fiction pipeline. Plan, draft, revise, export.
  gamehub-mod/    — Discord community moderation. Triage, audit watch, escalate.

Each profile contains:

SOUL.md — The agent's identity, personality, methodology, and output standards
skills/ — Domain-specific skills that give the agent its capabilities

Quick Start

1. Pick a profile

Browse the directories above. Each has a README.md explaining what it does and when to use it.

2. Create the profile in Hermes

hermes profile create <profile-name>

3. Copy the SOUL.md

cp profiles/<profile-name>/SOUL.md ~/.hermes/profiles/<profile-name>/SOUL.md

4. Install the skills

Copy the skills directory into your Hermes profile:

cp -r profiles/<profile-name>/skills/* ~/.hermes/profiles/<profile-name>/skills/

Or install individual skills from the Hermes skill registry:

hermes skill install <skill-name> --profile <profile-name>

5. Configure

Edit ~/.hermes/profiles/<profile-name>/config.yaml:

model: deepseek/deepseek-chat    # or anthropic/claude-sonnet-4, openai/gpt-4o, etc.
max_turns: 30
reasoning_effort: high
memory:
  enabled: true

6. Run

hermes chat --profile <profile-name>

Multi-Agent Fleet

The real power is running multiple profiles together. Senna acts as the front door — it routes your request to the right specialist.

You ask a question
    → Senna parses intent, identifies the domain
    → Routes to the right specialist profile
    → Specialist does the work
    → Results flow back through Senna to you

To set up a fleet:

Create the profiles you need
Copy SOUL.md and skills for each
Run Senna as your primary profile
Other profiles run as workers/orchestrators on their domains

Profiles: Orchestrators vs Workers

Type	Role	Examples
Orchestrator	Manages a domain, delegates to workers	code, creative, research, security, infra, mlops
Worker	Does focused work, reports up	finance, knowledge, homelab, media, social, communication, business, cyber-red, cyber-blue-cloud, cyber-blue-compliance, cyber-blue-forensics, cyber-blue-soc, educate, novel, gamehub-mod
Top Orchestrator	Routes across all domains	senna

Customization

Every SOUL.md is meant to be edited. The personality rubric, routing rules, and quality gates are starting points — tune them to your workflow.

Common customizations:

Change Report→Orchestrator to point at your actual orchestrator name
Adjust the PersRubric personality scores (0-100 per trait)
Add/remove skills for your specific needs
Modify DEFAULTS for your environment
Add Cron Duties sections for scheduled tasks

Guides

Getting Started With Hermes Profiles — Beginner orientation: profiles, model choices, fleet composition, PersRubric, token compression, educate usage
Research Profile Guide — Full walkthrough for setting up a research agent

What's NOT In Here

No personal paths, Discord channels, or API keys
No environment-specific config (your config.yaml stays in your ~/.hermes)
No runtime state (logs, databases, session history)

Everything here is portable. Copy what you need, customize what you use.

Third-Party Skills

This repo includes skills from the following open-source projects:

Anthropic Cybersecurity Skills

Author: Mahipal (mukul975)
Repo: github.com/mukul975/Anthropic-Cybersecurity-Skills
License: Apache 2.0
Used in: cyber-red, cyber-blue-* profiles
Description: 753 cybersecurity skills covering pen testing, forensics, threat intel, IR, cloud security, and more.

License

MIT for the profile definitions (SOUL.md files, READMEs, guides). Third-party skills retain their original licenses (see above).