code-yeongyu/oh-my-openagent

Codex QA verifies the omo Codex Light edition plugin in a real Codex app-server with an isolated CODEX_HOME and local mock model, so the user's real ~/.codex and API credentials are never touched. Use it whenever changes are made to the plugin, hooks, installer, or TUI.

Dead Code Removal scans a TypeScript project for unused code with TypeScript strict mode and LSP-verified safety, then removes it in parallel commits. Use it to clean up unused imports, constants, functions, and files without breaking builds.

Get Unpublished Changes compares the current git HEAD with the latest published npm versions and lists all unpublished changes grouped by release layer. Use it when you need to see what has changed since the last release to prepare changelogs or release notes.

Give Me Tips explains any senpi tip in depth—startup tips, working tips, and Tip: lines shown in the TUI—by querying the live tip list, checking what the user can see, and verifying the real feature code before answering.

Hyperplan orchestrates five adversarial category agents in team mode to cross-critique plans, distills only defensible insights, and hands them to the plan agent for executable formalization. Use when planning needs maximum rigor.

Hyperplan orchestrates a five-member adversarial multi-agent team that debates and critiques a plan to surface weak assumptions, then distills surviving insights and hands them to a planner for executable formalization. Use it when planning needs maximum rigor or when you want to expose blind spots and over-engineering.

Init Deep generates hierarchical AGENTS.md files for a project, scoring directories by complexity to decide where to place knowledge base files. Use it to create or update a knowledge base that guides AI coding agents.

Omomomo prints an easter egg message about Oh My OpenCode, a plugin that turns an AI agent into a multi-agent development team. Use it when the user triggers omomomo, about, or easter egg to share details about the plugin.

Onboarding walks first-time omo users through a guided six-lane conversation covering feature tour, migration help, session archaeology, value mapping, memory recording, and init-deep proposals. Use it during the initial omo session to set up the agent and record durable user facts.

Opencode QA verifies the opencode coding agent's CLI, HTTP server, plugin hooks/events, TUI, and session database using tested helper scripts. Use it to smoke-test, debug, or investigate opencode's behavior.

Pi Goal tracks long-running goals persistently, with Codex-style create, inspect, update, pause, resume, and complete commands. Use it when the user explicitly asks to set, continue, audit, pause, resume, complete, or inspect a goal.

Pre-Publish Review runs a 16-agent release gate before publishing npm packages. It detects unpublished changes, spawns per-change deep-dive and holistic review agents, and synthesizes release readiness and version bump recommendations. Use it when explicitly asked for a pre-publish review.

Publish automates the release process for oh-my-opencode: it triggers the publish workflow, verifies npm packages and release artifacts, drafts release notes, and posts the Discord announcement.

Security Research runs a parallel Team Mode audit with 3 vulnerability hunters and 2 PoC engineers to separate real exploitability from generic concern. Use it for security review, vulnerability research, exploitability audit, pre-release security check, and threat model validation.

Tech Debt Audit performs a thorough, file-cited technical debt audit across 9 dimensions using AST-grep, grep, language-native tooling, and optional CodeGraph. Use it for codebase health checks, architecture reviews, code quality assessments, or cleanup planning.

Triage fetches open issues and pull requests from a GitHub repository, spawns one background subagent per item, and writes evidence-backed reports to /tmp without mutating GitHub state. Use it when you need a read-only, permalink-verified analysis of an entire issue and PR backlog.

ULW Loop decomposes work into systematic, evidence-bound steps for durable, checkpointed goal execution, using subagents and manual QA channels. Use it for long-running, evidence-led delivery with git-tracked edits.

ULW Plan turns a vague or large request into a decision-complete work plan a downstream worker executes with zero further interview. It activates only when the user explicitly requests the ulw-plan workflow and never self-activates.

ULW Research orchestrates a cooperating AI team for maximum-saturation research: it scopes axes, fans out to sources, runs debate rounds, proves contested claims with code, and delivers a cited synthesis with charts and assets. Activate it when an explicit, exhaustive research request is made.

Ultrawork Mode activates a strict, evidence-driven coding workflow directive in omo-senpi when a prompt contains ultrawork or ulw, injecting it as a hidden context message ahead of the user's text.

Work With PR manages the full pull request lifecycle in isolated git worktrees, from implementation and manual QA through CI verification to merge and cleanup. Use it to split a task into atomic, independently mergeable PRs and build independent ones concurrently.