oh-my-hermes

Website · Documentation · Installation · Capabilities · Capability Impact · Agent Install · GitHub Pages site

[!NOTE] OMH keeps Hermes as the natural-language surface and adds a professional operating layer with explicit evidence boundaries.

Quick Start

Install the local command and managed skills:

curl -fsSL https://raw.githubusercontent.com/rlaope/oh-my-hermes/main/install.sh | sh
omh setup

Hermes skill tap path:

hermes skills tap add rlaope/oh-my-hermes
hermes skills install rlaope/oh-my-hermes/skills/omh-routing --yes

or ask Your AI Agent:

Hey Agent, Install this >> https://github.com/rlaope/oh-my-hermes <<

Update and health check:

omh update
omh doctor

Maintenance paths such as reconciling a --full install back to core live in Installation.

What OMH Adds

Mixture of Modes

OMH packages 92 installable workflow skills behind six human-readable capability families. The family is the front door; exact skill names remain available when a wrapper or operator needs precise control.

The full generated catalog, triggers, harnesses, and evidence rules live in Workflow Reference.

Highlights

Capability	Try it with	What it does
🧭 Clarify and plan	`ulw-interview` · `ulw-plan` · `omh-decide`	Turns an ambiguous request into explicit goals, constraints, tradeoffs, acceptance criteria, and a plan that can be handed off.
⚡ Build with leverage	`ulw-work` · `ulw-goal` · `ulw-team` · `ulw-loop`	Scales from fast parallel work to durable multi-step execution while keeping ownership, checkpoints, and verification visible.
🔬 Research and learn	`ulw-research` · `omh-best-practice-research` · `omh-research-brief`	Finds and synthesizes source-backed evidence with freshness, source-quality, and unresolved-uncertainty boundaries.
🛠️ Code and ship safely	`ulw-process` · `omh-code-review` · `ulw-qa`	Prepares executor-neutral coding work, then makes review, QA, CI, and merge claims depend on observed evidence.
🎨 Create polished deliverables	`omh-design-quality-gate` · `omh-materials-package` · `omh-deliverable-package` · `omh-img-summary`	Shapes websites, visuals, reports, decks, documents, PDFs, posters, and packages around content, taste, accessibility, and render-quality gates.
🧠 Remember and operate	`omh-memory-new` · `omh-memory-sync` · `omh-ops-observability-card` · `omh-doctor`	Keeps project memory review-first, surfaces operational readiness, and gives the next repair action without inventing provider or system state.
🔌 Connect without hiding boundaries	`omh-toolbelt-readiness` · `omh-external-connector-readiness` · `omh-agent-board`	Checks whether a needed tool, connector, or agent surface is really available before work depends on it, and keeps host load, tool use, and external-provider access separately observable.

Built For Real Work

OMH (Oh-My-Hermes) — Anyone can use hermes-agent professionally. The powerful intelligence harness for your AI Agent.

🧭 A stronger router, not a command dump. English, Korean, Japanese, Chinese, Spanish, French, German, and Hindi operator requests can be classified locally without a translation API. OMH returns the recommended family, skill, owner, next action, and what is still not evidence.

🤝 Better coding handoffs. OMH can include repository constraints, accepted scope, worktree and session-isolation guidance, locally available skills, acceptance criteria, review expectations, and verification gates. Codex, Claude Code, Hermes, and generic executors remain explicit owners rather than hidden defaults.

🎨 Quality-aware creation. Frontend, accessibility, image, report, slide, document, spreadsheet, PDF, poster, and shareable-package requests are routed through specialized production and QA guidance. A prepared brief is never presented as a generated or visually verified artifact.

🔍 Evidence before claims. OMH separates prepared intent, observed runtime events, and verified results. A handoff can be ready without claiming that an executor ran, a review passed, CI succeeded, a deployment completed, or a PR was merged.

🧠 Review-first project memory. OMH keeps project-memory candidates separate from approved records and recalls only reviewed, prepared context into future handoffs. It does not pretend to read or mutate opaque Hermes memory.

🔌 Provider-neutral operations. Metric, wiki, browser, image, video, and connector systems sit behind explicit external-provider contracts. OMH can validate and analyze supplied data without pretending that a provider was connected or called.

🏛️ Hermes-native, executor-neutral architecture. Hermes remains the chat, clarification, planning, research, and status surface. The selected executor owns implementation, while OMH supplies the local contracts, routing, memory, quality gates, and evidence boundaries around that work.

🧱 Local-first control plane. Core OMH routing, catalogs, manifests, and claim rules are deterministic local surfaces. External calls and provider access stay explicit integrations rather than hidden behavior inside the core.

Evidence Before Claims

OMH separates useful preparation from observed results:

State	Meaning
Prepared	A route, plan, prompt, artifact contract, or handoff is ready.
Observed	A wrapper or runtime recorded that an action or result occurred.
Verified	A matching test, review, served-surface check, or other required gate passed.

prepared_not_observed is not execution, provider access, artifact generation, review, CI, deployment, merge readiness, or a merge. Capability impact is reported across separate dimensions rather than collapsed into one marketing score. See Capability Impact.

Documentation

Development

For a source checkout:

PYTHONPATH=tests uv run python -m unittest discover -s tests -v
uv run python -m compileall -q src tests
uv run python -m omh.cli docs workflows --check
git diff --check

OMH is developed in the open as part of Team Art & Engineering. Follow @rlaope for project updates.

Oh-My-Hermes: Professional Agent Operating Layer

At a glance

README