ov-images/skills/hermes/SKILL.md
Full-featured standalone Hermes AI agent with AI CLIs, dev tools, DevOps tools, and ov. MUST be invoked before building, deploying, configuring, or troubleshooting the hermes image.
npx skillsauth add overthinkos/overthink-plugins hermesInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Full-featured standalone Hermes AI agent. No browser or desktop — designed for cross-container deployment alongside selkies-desktop (shared Chrome via BROWSER_CDP_URL) and jupyter (MCP notebooks via OV_MCP_SERVERS).
hermes:
base: fedora
layers:
- agent-forwarding
- hermes-full # hermes + claude-code + codex + gemini + dev-tools + devops-tools + ov + tmux
- dbus
| Layer | Purpose |
|-------|---------|
| agent-forwarding | SSH + GPG agent forwarding into container |
| hermes | AI agent with LLM auto-config, MCP, browser tools |
| claude-code | Anthropic Claude Code CLI |
| codex | OpenAI Codex CLI |
| gemini | Google Gemini CLI |
| dev-tools | bat, ripgrep, neovim, gh, direnv, fd-find, htop, podman-compose, etc. |
| devops-tools | AWS CLI, Scaleway, kubectx/kubens, OpenTofu, wrangler, bind-utils, jq, rsync |
| ov | Overthink CLI for in-container management |
| tmux | Terminal multiplexer for persistent sessions |
| dbus | D-Bus session bus |
ov image build hermes
ov config hermes -e OLLAMA_API_KEY=your-key # or OPENROUTER_API_KEY
ov start hermes
ov shell hermes -c "hermes chat"
The hermes entrypoint performs a single-phase, first-start-only auto-configuration of LLM providers and MCP servers. Guarded by # ov:auto-configured sentinel in config.yaml. To reconfigure: delete /opt/data/config.yaml and restart. API keys are synced to .env on every start (handles rotation).
LLM provider based on which env var is set:
| Priority | Env var | Provider | Default model |
|----------|---------|----------|---------------|
| 1 | OLLAMA_HOST | Local Ollama | qwen2.5-coder:32b |
| 2 | OLLAMA_API_KEY | Ollama Cloud | kimi-k2.5:cloud |
| 3 | OPENROUTER_API_KEY | OpenRouter | qwen/qwen3.6-plus:free |
Override the default model with HERMES_MODEL env var. Additional env_accepts:
| Variable | Description |
|----------|-------------|
| TELEGRAM_BOT_TOKEN | Telegram bot token for messaging |
| SLACK_BOT_TOKEN | Slack bot token |
| DISCORD_BOT_TOKEN | Discord bot token |
# Ollama Cloud
ov config hermes -e OLLAMA_API_KEY=your-key
# OpenRouter with custom model
ov config hermes -e OPENROUTER_API_KEY=sk-or-xxx -e HERMES_MODEL=google/gemini-2.5-flash
# Local Ollama (OLLAMA_HOST auto-injected by ov config ollama --update-all)
ov config hermes
# Or via .env file in the data volume
ov shell hermes -c "vi /opt/data/.env"
The data volume (/opt/data) persists: sessions, skills, memories, logs, config, and .env.
Deploy alongside provider containers for full functionality:
# 1. Deploy selkies-desktop (provides BROWSER_CDP_URL)
ov config selkies-desktop
ov start selkies-desktop
# 2. Deploy jupyter (provides jupyter MCP server)
ov config jupyter --update-all
ov start jupyter
# 3. Deploy hermes (consumes both)
ov config hermes -e OLLAMA_API_KEY=... --update-all
ov start hermes
Hermes receives:
BROWSER_CDP_URL=http://ov-selkies-desktop:9222 — controls desktop ChromeOV_MCP_SERVERS=[{"name":"jupyter","url":"http://ov-jupyter:8888/mcp"},{"name":"chrome-devtools","url":"http://ov-selkies-desktop:9224/mcp"}] — notebook manipulation + browser DevTools MCPWhen co-deployed with services that declare mcp_provides (e.g., jupyter), hermes auto-discovers and connects to their MCP servers at first start. The OV_MCP_SERVERS JSON env var is injected by ov config and the entrypoint writes the servers into config.yaml under mcp_servers:.
# Verify MCP connection
ov shell hermes -c "hermes mcp list" # Shows registered servers
ov shell hermes -c "hermes mcp test jupyter" # Tests connection (expects 13 tools)
/ov-layers:hermes-full — Metalayer composition details/ov-layers:hermes — Core agent (env_accepts, browser dispatch, LLM config)/ov-layers:chrome — Provides BROWSER_CDP_URL (from selkies-desktop)/ov-layers:chrome-devtools-mcp — Chrome DevTools MCP server on port 9224 (from selkies-desktop)/ov-images:hermes-playwright — Hermes with local Playwright Chromium/ov-images:selkies-desktop — Desktop with Chrome (cross-container browser provider)/ov-images:jupyter — JupyterLab with MCP (cross-container MCP provider)ov status hermes
ov service status hermes # hermes: RUNNING
ov shell hermes -c "hermes --version"
ov shell hermes -c "claude --version"
ov shell hermes -c "codex --version"
ov shell hermes -c "gemini --version"
ov shell hermes -c "ov version"
ov shell hermes -c "echo BROWSER_CDP_URL=\$BROWSER_CDP_URL"
ov shell hermes -c "echo OV_MCP_SERVERS=\$OV_MCP_SERVERS"
ov shell hermes -c "hermes mcp list" # Should show chrome-devtools, jupyter
Latest ov test hermes run: 50 passed, 0 failed, 0 skipped.
Covers all 4 AI CLIs at ${HOME}/.npm-global/bin/{claude,codex,gemini}
hermes at ${HOME}/.pixi/envs/default/bin/hermes, plus
dev-tools (rg, bat, gh, fastfetch, nvim, htop) and devops-tools (aws,
scw, kubectx, kubens, tofu, jq). Deploy-scope: pipewire + hermes
services up, /opt/data volume mounted. Uses supervisorctl pid for
liveness (hermes-whatsapp is autostart=false — see /ov:test Gotcha #4)./ov-layers:hermes-full, /ov-layers:hermes, /ov-layers:claude-code,
/ov-layers:codex, /ov-layers:gemini, /ov-layers:dev-tools,
/ov-layers:devops-tools/ov:test — declarative testing framework + supervisord gotchas/ov:config — OV_MCP_SERVERS auto-discovery + secret provisioning/ov:mcp — verify each MCP server in OV_MCP_SERVERS is actually alive before debugging hermes tool-call failures; ov test mcp list-tools <provider-image> shows exactly what hermes will see/ov-images:selkies-desktop — companion for shared browser (CDP)/ov-images:jupyter — MCP notebook tools auto-discoveredMUST be invoked before building, deploying, configuring, or troubleshooting the hermes image.
/ov:image — image family umbrella (image: entries in overthink.yml, build/validate/inspect/list)/ov:build — build.yml vocabulary (distros, builders, init-systems)tools
Use when authoring or modifying a charly PLUGIN — a candy with a `plugin:` block that contributes Providers (verbs/kinds/deploy-targets/steps/builders/commands), its own CUE schema, builtin (compiled-in) or external (out-of-tree git repo). Covers the unified Provider model, the per-plugin CUE-schema contract (single source → Go params for dev + schema-over-Describe RPC for runtime), the SDK, and the loader.
tools
The CUE data-validation / configuration CLI (cue), pinned to v0.16.1. Use when working with the cue candy, installing the cue binary into a box or onto a target:local dev host, or running the offline schema-vendoring pipeline that feeds charly's egress validation.
tools
CUE EGRESS validation — validating (and, where it adds value, generating) the config files charly WRITES to a system BEFORE the bytes hit disk. MUST be invoked before working on charly/egress.go, the vendored schemas under candy/plugin-egress/egress-schemas/vendor/, the ValidateEgress / registerVendoredEgressKind path, the offline `task cue:vendor` pipeline, or adding an egress schema for any written artifact (cloud-init, k8s manifests, traefik routes, runtime config, install ledger, systemd/quadlet units, ssh_config, libvirt XML).
tools
Kubernetes cluster-probe declarative check verb — the `kube:` check verb (nodes, pods, ingress, storage class, addon health, apply/delete, and arbitrary resource GETs) served out-of-process by the candy/plugin-kube plugin (vendored client-go; no external kubectl required).