🌓

Skills & Commands

The MultiTerminal plugin ships 15 skills — slash commands that orchestrate the agent workflow — and 9 specialist agents that the workflow skills dispatch for review, debugging, and session work. Each skill is invoked as /<name>; each agent is spawned automatically by the pipeline and project-management skills.

What skills are

A skill is a packaged instruction set Claude Code loads on demand when you type /<name> (or when the model recognizes a matching intent). Each lives in skills/<name>/SKILL.md inside the plugin, with a frontmatter name/description the model uses to decide when to invoke it. Intent matching means you don't have to type the slash command at all — saying "act as PM" or "dispatch these tickets" pulls in /program-management, and "verify the build" pulls in /verifier-multiterminal, because those phrases match the skill's description. Skills are instructions for the agent to execute — they drive the Bash, PowerShell, MCP, and Task tools — not standalone scripts. The 15 skills fall into three groups: the workflow skills that run the kanban lifecycle, the setup/onboarding skills that bootstrap projects and connectivity, and the utility/intel skills that handle profiles, briefings, and verification.

Workflow skills

These five drive the day-to-day task lifecycle — from picking work, through single- and multi-ticket coordination, to the post-coding review gates.

Command	What it does	When to use
/session-start	Lightweight startup menu. Establishes identity, registers the session in the lifecycle pipeline, reads active context, and presents quick choices (continue task, new task, pick task, just chat). No heavy loading until you pick.	Auto-runs at the start of every session; the entry point that routes you into the heavier workflows.
/kanban-task	Orchestrates a single ticket's full lifecycle: auto-detects current state, guides planning/coding/testing cycles, enforces completion gates, and writes continuation notes for clean session handoffs.	Working one ticket end-to-end, resuming after context ran out, or checking your current task state.
/project-management	Orchestrates the full development lifecycle — task dashboard, planning, team assembly, code review, build verification, and testing. Tiered SMALL/MEDIUM/LARGE workflows scale ceremony to task complexity.	Picking up new work, managing agents, or coordinating multi-step development (options 2–3 of the start menu).
/program-management	Coordinates a multi-ticket program as a PM — sequences 2+ related tickets across agents, runs quality-gate cycles, and triages findings across cycles. Distinct from /kanban-task's single-ticket scope.	"Act as PM" / "dispatch these tickets" — any batch of 2+ related tickets coordinated together.
/pipeline	Runs the complete post-coding review pipeline — verifier, code-reviewer, security-auditor, debugger, and a cross-model adversary — each gate dispatched to Claude, Codex, or off per the saved topology. Verifier runs first (must pass build), then the rest in parallel; failures cycle back as coding items and it loops until all gates pass.	"Run the pipeline" / "full review" / "run all gates" once coding is complete.

Setup & onboarding skills

These five bootstrap projects into MultiTerminal, stand up the remote/phone connectivity stack, and rebuild context after a clear.

Command	What it does	When to use
/new-project	Interactive onboarding wizard with two modes: Create New (scaffold folders from an archetype — Clarion COM, WebView2, Clarion App, MT Feature, Generic C# — init git, register) or Add Existing (point at a folder, auto-discover details, confirm, register).	Starting a brand-new project, or registering an existing folder so MultiTerminal can track it.
/multiterminal-addproject	The lower-level path: creates .claude/project.json in the current directory and registers it in the central MultiTerminal registry.	Quickly registering the current working directory as a project without the full wizard.
/multi-connect-setup	Guided, mostly-automatic phone connectivity via Tailscale: detects/installs Tailscale, brings the node up (one manual browser login), publishes the loopback gateway over HTTPS with tailscale serve, writes the hostname back into the Multi-Connect config, and prints the phone URL. Windows only.	Connecting a phone to MultiTerminal or configuring remote/Multi-Connect access.
/start-servers	Starts the remote-access server stack — the HTTP bridge that relays prompts, questions, and permission requests to your phone (see Multi-Connect), Caddy (HTTPS), and optionally the TTS server — checking which ports are already listening and only starting what's missing.	Beginning a remote phone-access session when the user needs phone access.
/reload-context	Rebuilds session context after /clear — replays what the SessionStart hooks inject: terminal identity, ACTIVE-CONTEXT.md, kanban tasks, active task detail, and the last session recap.	After a /clear, or "reload context" / "what was I working on".

Utility & intel skills

These five cover profiles, the daily intelligence/briefing pipeline, the Claude Code internals search, and build verification.

Command	What it does	When to use
/profile	A friendly front-end for updating your team-member profile — personal info, skills, interests, avatar — auto-detecting the calling agent from MULTITERMINAL_NAME so you never specify identity manually.	Filling in or updating your team profile without raw MCP calls.
/daily-digest	Runs the bundled DailyDigest fetch script — a data-collection helper that gathers raw AI/Claude-ecosystem data from GitHub, Reddit, HN, and the web — then summarizes the results into an actionable morning briefing.	"Daily digest" / "morning briefing" / "what happened today in AI".
/daily-intel	A four-phase intelligence pipeline: digest triage (action items → suggestion tasks), version watch (new Claude Code CLI releases via npm), KB update (fetch + append changelog), and an Owner push notification with the summary. Can be cron-triggered daily.	Daily automated intel, or manually to process the digest and check for CLI updates.
/claude-code-search	Searches the Claude Code internals knowledge base — architecture, tools, commands, state management, and hidden features distilled from a source-code analysis.	Looking up how Claude Code itself works (tools, slash commands, the query engine, state).
/verifier-multiterminal	Verifies MultiTerminal code changes by building the project, validating MCP tool operations, and checking REST API health. Constrained to a build/inspect tool set.	"Verify the build" / "check if it works" — after coding, before code review.

Specialist agents

The plugin's agents/ directory defines 9 specialist subagents. These aren't invoked by slash command — they're dispatched automatically by the workflow skills: /pipeline fans out the review gates, and /project-management / /program-management spawn the planning, design, and session agents at the right lifecycle step. Most run on Opus; each ships with a constrained tool set scoped to its job. (The folder also contains report-template.html, a shared HTML report template the review agents render into — it is a template, not an agent.)