Commands

You never need a command — plain language in any language works; Mops parses intent and asks when ambiguous. Every command is namespaced (/multica-ops:<verb>) because that is how plugin commands work: the prefix is what stops two plugins colliding over one verb. There is no bare /<verb>, and outside Claude Code there are no slash commands at all.

Three doors accept the same opening sentence, and the order is fixed. Something is broken goes to bug when it is a defect in the product and reproducible — that lane jumps the queue. Someone told us something goes to mops feedback, which triages into accept · decline with a reason · duplicate · snooze. report is for everything else that went wrong, including when you cannot tell which of the two it is — it decides from the evidence and may route into bug itself, but never skips feedback's triage on the way.

Nineteen doors, and the palette is the point. A verb earns a door when it is its own flow, reached by name, repeatedly — never as a synonym for another door, and never as a phase inside one. Everything else is a sentence to the dispatcher, which costs nothing to keep and is listed further down so the capability stays findable.

/multica-ops:mops <anything> is that dispatcher: free text, any language, any of the flows below (/multica-ops:mops brand, /multica-ops:mops move crossfeed to the next release). Where column: 🖥️ console (heavy/machine/interactive) · 🏢 Mops in Multica (presence/async) · ⇆ either.

The nineteen

Command	Where	Routes to
/multica-ops:mops <anything>	⇆	the front door — free text, or nothing at all: day zero, then read what you have and route. Also the way to every flow in Everything else below
/multica-ops:init	🖥️	bootstrap from zero: day zero (BOOTSTRAP §0) → three routing questions → shaping → interview → stand up. Full procedure: FLOWS
/multica-ops:join	🖥️	join an existing setup — audit + interview delta, then gaps → fixes; reconcile an existing Mops in Multica. Full procedure: FLOWS
/multica-ops:import	🖥️	bring a backlog in from another tracker (Linear, Jira, GitHub Issues, Trello, CSV): extract → show the mapping → create parents then children, unassigned, with source_id in metadata so a rerun resumes. Its own door because declaring a crossing turns on two rules — the mapping is shown before anything is written, and the imported text is treated as untrusted. PLAYBOOKS
/multica-ops:quick <task>	🖥️	the quick-job shape, direct — 3 questions (deliverable · repo · language), 1–2 agents, build → review, none of the company machinery; whatever it outgrows is added later
/multica-ops:consult [who] <question>	⇆	a question, not a build: answered as an advisor with zero standing footprint — no workspace, issue, team or doc created, nothing persisted unless you ask. Address Mops (default), any team agent, an expert, or the theatre. Nothing will be created is worth being able to demand. Full flow: FLOWS
/multica-ops:status	⇆	the console read, opening with a countable state line ("3 need you · 2 running · 4 closed since Tuesday"), then two parts that differ in kind: needs you — open requests with their age and what the wait costs, staying until answered — and happened — events since your last look, ageing out by themselves. Drift is reported separately. Plus spend and ripe deferrals; offered unprompted when you return after time away
/multica-ops:next	⇆	start the next feature from ROADMAP.md — the manual flow's main button
/multica-ops:feature <text>	⇆	add a feature mid-flight — raw description is fine: the conductor grills you into a spec → ICE against the backlog → proposed release slot → your approval → queued. Too fuzzy to spec? It offers discovery first
/multica-ops:ship [release]	🖥️	the go-live step — whatever "live" means here: ship code, launch a flavour, publish an episode, send the batch. Gates green → release → notes → tag → announce (deploy and announce are outward → owner-confirmed); writes the cost/effort ledger and marks it shipped
/multica-ops:bug <text>	⇆	the urgent lane that jumps the queue — a defect, recall or correction. Minimal spec → Build (first sub-step: a reproduction / pass-fail signal, before any fix) + Review, owner notified; not ICE-prioritized
/multica-ops:recover	🖥️	revive after limits — rerun interrupted work, revive marker-less cancels; a rerun resumes, it does not restart
/multica-ops:hire <role|person>	🖥️	add to the team — Mops asks agent or real person. Agent → the role-builder. Person → workspace member invite (owner-confirmed, outward) → access and review checkpoints → recorded in TEAM.md
/multica-ops:fire <agent|member>	🖥️	offboard — surface the risk first (open issues, squad leadership, sole-owner skills, held checkpoints) → reassign → archive. Removing a real member is the owner's own action in the Multica app, and Mops says so
/multica-ops:audit	🖥️	health and opportunities: token burn, limit-killed runs, tier misfits, stalled work, hygiene, mention cycles, secrets, design-system drift · process improvements · integrity checks (facts past their check-date, gates where author and reviewer coincide, edits to locked surfaces, unsourced ICE scores) · the waste slices (PLAYBOOKS → ledger) · the gates table (each enforced_by still true — PLAYBOOKS → Gates). Pulls in the health sweep. Output: finding → recommendation → impact
/multica-ops:report <text>	🖥️	something went wrong, or made the work harder — and you do not have to know whose defect it is. Decided from the evidence: a defect in your product goes to the urgent lane (bug); friction in your workspace becomes a line in _ops/FIELD-NOTES.md, swept at the next audit, an issue on the second occurrence with both named; friction in this skill or Multica is packaged from evidence, de-identified, and written to a file outside your repository with its path said and the routes named. You post it, never us
/multica-ops:upgrade [skill|all]	🖥️	the one command for getting current, four layers: this skill's bytes (Mops runs it; content applies on next read — a restart only for a new or renamed command or a hook), the workspace migration, imported skills, and the CLI locally and on runtimes — reported, not run, and only when active_task_count is 0. Re-screen → dry-run → back up → apply → reconcile → verify/rollback, then the plain-language tour of what changed
/multica-ops:skill <create|import|optimize|release>	🖥️	the toolkit's lifecycle: create (a routine seen twice), import (screen for danger and injection → trim → attach with provenance), optimize (fail-closed compression; NOT_COMPRESSIBLE is a valid answer), release (proven → de-identified → its own repo → re-imported). PLAYBOOKS
/multica-ops:cli <command>	🖥️	raw Multica CLI escape hatch — run or explain any multica … command directly, no methodology assumed. Backed by REFERENCE §10; destructive and outward commands still need owner sign-off

Everything else is a sentence

These are flows, not commands — every one is reached with /multica-ops:mops <verb> or plain language in any language. They lost their door because a door is paid for in every session by every agent, and these are not the ones people reach for by name.

Planning and the product

• roadmap — view or rebuild the release plan, re-run ICE, release surgery (cut, extend, reprioritize). "show me the roadmap"
• move · drop — one feature between releases or to the backlog · cancel with a reason (or park it). "push crossfeed to the next release"
• map — the product as a map, not the repo as a tree: _ops/MAP.md, the things and the moves in the product's own words, current state only, ending honestly with not mapped yet. "map the product"
• measure — pull the success metrics set at discovery, compare to target, file a Learn item, record the ledger. "did it move the metric?"
• research — point research without a feature; cited findings land in _ops/research/, each carrying the date it was true. "look into how competitors price this"
• discovery — spin up a fuzzy idea: research, competitors, team brainstorm → proposal. "I'm thinking about onboarding"
• feedback — log an incoming signal, then triage into accept · decline with a reason · duplicate · snooze; declining is a normal, cheap outcome. "a customer says export is confusing"
• process — discover a craft's real process, then a tool per step by function. "how should we run this kind of work?"

The team and its shape

• team · squad — the roster (🎭 theatre entities excluded from headcount) · create and reshape squads and their routing. "who's on what?"
• update — reconfigure an agent or a person: instructions, skills, squad, tier, or a project lead. "make the copywriter own localization too"
• crew — crew mode: executors and gates, no conductor, the owner is the PM. The default offer after an import when no conductor stands. "just execute, I'll prioritise"
• module — toggle an opt-in module (resident Mops, design system & brand, experts, personas, Design QA, social). "turn on personas"
• audience — segments, ICP, personas as documents, staged proto or validated, with live participants asked about explicitly. Full theatre: MODULES
• validate — an artifact past the validators; synthetic and live reactions counted separately, synthetic verdicts direction-only, and the gate handed back to whoever owns it
• brand — create or evolve the brand book; an existing brand is audited, not rebuilt

Governance and cost

• access · reviews — what a member may direct Mops to do · which flows need a named human's sign-off. "anna can start features but not spend"
• budget — the envelope: amount, period, currency, credits with their expiry cliffs; it shapes advice, not only caps it. "$50 a month"
• autonomy · pace — who starts the next feature and whether hiring needs a yes · how wide to fan work out. "go non-stop" · "faster"
• autopilot — scheduled or webhook automations; a webhook URL is all four owner-gated kinds at once

Operations

• health — the silent-failure sweep: runtimes, integrations, tokens, free-tier headroom, branch protection, daemon, limits. Folded into audit
• sync — two-way reconcile of drift the fingerprint caught, attributed before it is asked about
• start · stop — the local daemon. "stop everything"
• workspace — list or switch the active workspace; Mops confirms which company it is acting on
• switch — reassign runtime, model or thinking-level per agent, or an assisted whole-provider migration
• connect — integrations: MCP registries first, study the tool, wire access, register in _ops/TOOLING.md

In the workspace the user talks to Mops in Multica (subtitle "Executive Advisor · resident") — no commands needed, plain chat; it answers status-style questions natively and, for anything 🖥️, points back to the console.