rlaope/oh-my-hermes

Accessibility Audit reviews UI surfaces for WCAG 2.2 AA evidence, including keyboard reachability, focus order, screen-reader behavior, target size, contrast, and reflow. Use it when you need an accessibility pass or audit before claiming compliance.

Adding A Skill explains how to register a new installable skill in OMH, update routing and prompts, and regenerate the generated docs. Use it when adding a skill and you need to satisfy the repo's required checks and fixtures.

Agent Board coordinates multiple Hermes profiles or agents with task, handoff, heartbeat, blocker, and completion states. Use it for board-shaped collaboration when you need structured status tracking and evidence boundaries.

Agent Debug captures a stuck, looping, or drifting agent run and identifies the likely failure pattern. Use it when an agent is burning tokens, repeating tools, or failing on recoverable assumptions and you need the smallest safe recovery step.

Agent Evaluation compares Codex, Claude Code, Hermes, or other executors on reproducible tasks using quality, cost, time, tool, and evidence metrics. Use it to choose a runtime or benchmark agent output fairly.

Agent Ops Review helps managers inspect AI-agent progress, blockers, quality gates, and throughput levers. Use it when you need a structured status view of agent work and its next actions.

Ai Slop Cleaner guides behavior-preserving cleanup of duplicated, dead, or AI-generated code. Use it when you need regression-gated refactoring that keeps observable behavior unchanged.

Ask routes a question to an external advisor such as Claude or Gemini when configured. Use it for explicit consultation requests that need a second opinion before proceeding.

Automation Blueprint turns recurring Hermes requests into a scheduled operations blueprint with cadence, delivery target, silence policy, and context-chain selection. Use it when you need a cron-like plan without claiming runtime execution or delivery.

Autoresearch Goal runs validator-gated, durable research loops with explicit evidence boundaries and source-gap tracking. Use it when a research task needs checkpoints, critic checks, and a source-backed handoff.

Best Practice Research gathers official or upstream guidance for a specific technology and compares it to the user’s question. Use it when you need source-backed best practices from docs before choosing or implementing an approach.

Browser Operator prepares a structured, evidence-bounded browser task card for opening URLs, clicking, logging in, and filling forms. It is used when browser actions need auth gates, confirmation gates, and observed-trace boundaries instead of ad hoc narration.

Build Failure Triage inspects failing build, typecheck, lint, test, CI, and DCO checks to produce a minimal, evidence-backed fix handoff. Use it when you need to classify failures and plan the smallest safe remediation or rerun path.

Buzz connects Hermes to the native Buzz gateway for setup, media delivery, or self-hosted relay diagnostics. Use it when you need verified relay evidence or need to troubleshoot a Buzz connection without inventing delivery proof.

Cancel ends an active Hermes workflow state cleanly when the user says cancel, stop, or abort. Use it to clear state and return a safe stop summary without claiming unobserved execution.

Capability Toggle changes whether an OMH capability family is offered, such as memory or coding orchestration. Use it when you want to enable or disable a family without uninstalling OMH.

Codebase Onboarding creates a repo map, reading path, glossary, risk map, and first-task runway for unfamiliar codebases. Use it when you need to understand a new repository before planning implementation.

Codegraph Refresh refreshes local code intelligence, summarizes repo structure, and prepares task-scoped handoff context before planning or implementation. Use when codegraph data is stale or a fresh navigation summary is needed without claiming execution evidence.

Connector Operator prepares evidence-bounded external app actions for services like email, Slack, Jira, Notion, and CRM tools. Use it when you need a gated connector workflow for sending, creating, or updating provider state.

Content Operator prepares structured, evidence-bounded workflows for drafting, rewriting, summarizing, translating, and release-note or newsletter copy. Use it when you need publish-ready content with audience, tone, source, review, and hallucination gates.

Context Budget Review plans compact context for long agent work, including token and cost budgets, summarization checkpoints, and overflow recovery. Use it when context size or drift could affect research, coding, or review tasks.

Cto Loop runs a leadership operating loop for roadmap, technical tradeoffs, delivery risk, release readiness, and follow-up handoffs. Use it when a request needs CTO/PM-style synthesis across product, engineering, QA, security, and ops.

Curriculum Design turns a learning goal into a teachable sequence with objectives, scope, assessments, and learner constraints. Use it when planning lessons, modules, or evaluation criteria for a course or enablement program.

Hermes Achievements summarizes unlocked badges, tiers, recent unlocks, and progress from local hermes-achievements artifacts. Use it when you need an achievements summary or badge status from observed plugin data.

Hermes Code Review runs a bug-first, evidence-grounded review of code, diffs, and claims. Use it when you need ranked findings, test gaps, and a clear ship or no_ship verdict.