⚡ Hermes Agent Skills

Production-grade skills + domain agents for Hermes Agent — self-evolving, persona-aware, slash-command native.

  DEFINE           BUILD          VERIFY           SHIP           EVOLVE
 ┌────────┐     ┌────────┐     ┌─────────┐     ┌────────┐     ┌──────────┐
 │  Spec  │ ──▶ │  TDD   │ ──▶ │  Debug  │ ──▶ │ CI/CD  │ ──▶ │ Curator  │
 │Design  │     │  Code  │     │  Gate   │     │ Deploy │     │ Persona  │
 └────────┘     └────────┘     └─────────┘     └────────┘     └──────────┘

✨ Features

🧬 Self-evolving — Skills aren't static. The built-in EvolutionEngine tracks 5 health dimensions (usage frequency, success rate, corrections, freshness, command validity), scores each skill, and surfaces improvement suggestions through Hermes's native /curator system.
🎭 Persona-aware — Drop a SOUL.md in your Hermes config and every code-generating skill adapts — naming, comments, architecture. Your agent writes code that looks like you wrote it.
🔗 Hermes-native — Every skill leverages Hermes-specific tools: delegate_task, browser+terminal+vision coordination, persistent memory, /curator, cronjob, webhook.
📐 Open standard — All SKILL.md files follow the Agent Skills Open Standard. Validated by the bundled SkillValidator.

🧪 Hermes v2 Beta — Task-First Interface

v1.1.0 remains stable and supported alongside v2.

Hermes v2 adds a natural-language frontend: say what you want, the system handles roles, skills, and constraints automatically.

# v2 Beta — natural language
python -m hermes_v2.cli "帮我发布项目"
python -m hermes_v2.cli "创建一个 FastAPI 项目"
python -m hermes_v2.cli "修复这个错误"

# v1 — always available
hermes-skill validate skills/

Three modes: Beginner (auto) · Advanced (choose persona) · Expert (pure v1). See Hermes v2 MVP docs.

v1 is unchanged. v2 is additive — all existing skills, CLI commands, and workflows work as before.

📦 Workflow Skills (8 SKILL.md)

Skill	Phase	Description	Hermes-specific
`requirement-analyzer`	Define	Structured multi-turn requirement extraction	Cross-session persistent memory
`spec-driven-dev`	Define	Seven-section PRD before any code	`/skills` pipeline chaining
`test-driven-dev`	Build	RED-GREEN-REFACTOR + test pyramid	`delegate_task` parallel testing
`debugger-coordinator`	Verify	Multi-modal debugging (5-step method)	`browser`+`terminal`+`vision` coordination
`code-quality-guardian`	Verify	Six-axis quality gate	`patch` auto-fix + `/curator` tracking
`cicd-orchestrator`	Ship	GitHub Actions workflow generation	`cronjob` + `webhook` triggers
`skill-curator`	Evolve	Collect → Analyze → Propose → Execute	Direct `/curator` integration
`persona-aware-coding`	Evolve	SOUL.md-driven style adaptation	Native Hermes identity system

🤖 Domain Agents (from HermesHub)

Plug-and-play specialist agents — each is a self-contained bundle of persona, memory, and dispatchable Python skills.

Agent	Domain	Skills	Install
Python Pro	Python 3.11+	Code Review · Performance Profiling · Test Generation · Scaffolding · Type Checking	`hermes skills install Ow1onp/hermes-agent-skills/skills/agents/python-pro`
DevOps SRE	Infrastructure	CI/CD Pipeline · Docker Optimization · K8s Deploy · Log Analysis	`hermes skills install Ow1onp/hermes-agent-skills/skills/agents/devops-sre`

Domain agents use the SCHEMA + handler() pattern directly compatible with Hermes's tool dispatch system. Skills activate only when your intent matches — no context pollution.

🚀 Quick Start

# Recommended: install via hermes skills tap
hermes skills tap add Ow1onp/hermes-agent-skills
hermes skills browse
hermes skills install Ow1onp/hermes-agent-skills/skills/define/requirement-analyzer
hermes skills install Ow1onp/hermes-agent-skills/skills/agents/python-pro
hermes skills install Ow1onp/hermes-agent-skills/skills/agents/devops-sre

# Load a workflow skill
/skill requirement-analyzer

# Load domain agents
/skill python-pro
/skill devops-sre

hermes skills tap add Ow1onp/hermes-agent-skills points Hermes at the repo's skills/ tree. Workflow skills live under phase folders such as define/ and build/; domain agents are exposed through installable wrappers under skills/agents/. The top-level agents/ tree remains the source bundle for agent persona, memory, and handler code. Manual copying to ~/.hermes/skills/ is only a local debugging fallback.

Short-name installs such as hermes skills install python-pro may work once Hermes's registry/index resolver has refreshed, but the full GitHub identifiers above are the deterministic install path.

🔗 Hermes Integration

# Self-evolution engine
from hermes_agent_skills import EvolutionEngine
engine = EvolutionEngine()
engine.record_task(...)
suggestions = engine.analyze()  # detects stale skills, recurring patterns

# Persona-aware coding
profile = SoulReader().read("~/.hermes/SOUL.md")
print(profile.get_code_prompt_hint())

📂 Project Structure

hermes-agent-skills/
├── skills/                  # SKILL.md workflow skills (Agent Skills standard)
│   ├── define/ · build/ · verify/ · ship/ · evolve/
│   └── agents/              # Hermes-installable domain agent wrappers
├── agents/                  # Domain agents (from HermesHub)
│   ├── python-pro/          # SKILL.md + persona.md + memory.md + 5 handlers
│   └── devops-sre/          # SKILL.md + persona.md + memory.md + 4 handlers
├── src/hermes_agent_skills/ # Python library
│   ├── validator.py         # SKILL.md validator
│   ├── evolution.py         # Self-evolution engine
│   └── soul_reader.py       # SOUL.md persona parser
├── src/cli/                 # hermes-skill CLI (Typer)
├── tests/                   # 46 tests, all passing
├── scripts/                 # Utility scripts
└── .github/workflows/       # CI/CD (lint + test matrix + validate + security)

📄 License

MIT © Ow1onp

🤝 Contributing

Skills are plain Markdown. Issues and PRs welcome.

git clone https://github.com/Ow1onp/hermes-agent-skills.git
# Create your skill at skills/<phase>/<name>/SKILL.md
pytest tests/test_validator.py -v
# Open a PR

Hermes Agent Skills

README