メニュー
Answer questions about Main Branch and Claude Code. Use when: user asks how/what/why questions, is confused about the business folder, CLI, skills, setup, errors, workflow, or what to do next.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | mb-help |
| description | Answer questions about Main Branch and Claude Code. Use when: user asks how/what/why questions, is confused about the business folder, CLI, skills, setup, errors, workflow, or what to do next. |
| loops | ["sense","decide"] |
Answer questions, troubleshoot issues, explain philosophy, suggest next steps.
CLI facts first: For "what should I do?", setup, provider, skill wiring, or
troubleshooting questions, run mb status --json --peek and use mb doctor
for repair facts before giving prose-only advice.
mb status --json --peek and/or mb doctor repair --plan before relying on it.| Keywords | Reference |
|---|---|
| Terminal, drag files, cd, business folder vs site folder | terminal-basics.md |
| Business folder, CLI, skills, linked site folder, system model | system-model.md |
| Philosophy, why, compound, passive memory | philosophy.md |
| /mb-think, research, decide, codify | the-think-cycle.md |
| /mb-bet, bet, hypothesis, deadline, public narration, bet vs offer | skills-guide.md |
| Work continuity, where left off, GitHub issues | work-continuity.md |
| Error, command not found, MCP, Apify setup, GitHub issue | troubleshooting.md |
| Provider readiness, GitHub setup, Cloudflare, Google Workspace, Meta Ads, Apify | provider-readiness.md |
| Getting started, setup | Route to /mb-setup or /mb-start |
| Which skill, when to use, list skills, what skills do I have, skills inventory, what can you do, which skills are we using | skills-guide.md |
| Pre-publish review, UI pass, design pass, design critique, review the page before publish | Route to /mb-skill-review (see skills-guide.md) |
| Create skill, Notion export, custom | skills-guide.md |
| Migrate from GPT, ChatGPT | gpt-migration.md |
| Reels, TikTok, organic, /mb-organic | organic-help.md |
| Skool, community | skool-help.md |
| Wiki, atomic notes, publish, WikiLinks | Route to /mb-wiki |
| Launch readiness, Cloudflare, Wrangler, Shopify, Resend, smoke test, form test, booking link | Route to /mb-site; run mb launch check before launch-readiness advice |
| Site, landing page, lander, minisite, website, about page video, deploy site, publish site, conversion endpoint, concept variations | Route to /mb-site (see skills-guide.md for the full one-flow walkthrough) |
| Done, wrapping up, closing, end session, end of day | Route to /mb-end |
| multi-offer, product ladder, offers, switch offer | Multi-Offer FAQ (below) |
| Better outputs, quality, what next | making-outputs-better.md |
| Content strategy, pillars, platforms, newsletter, channel strategy, account strategy, founder voice, weekly content plan | content-strategy-help.md |
| Subagents, parallel, agents, context window, tokens | working-with-agents.md |
| workspace-isolated tools, sandboxed workspace, skills not showing | workspace-setup.md |
| Contribute, contributor | becoming-contributor.md |
/mb-think, /mb-setup, /mb-ads, /mb-site, /mb-organic)mb issue draft for reproducible Main Branch friction, but never submit issues for the operator.| Question | Answer |
|---|---|
| Is this a bet or an offer? | If you're testing whether a direction should continue, it is a bet in bets/. If it is something the business sells or may sell repeatedly, it is an offer. A live idea can be both: open the bet first, then create or update the offer only when the operator wants durable sellable truth. |
| Start Claude in a folder? | Open the business folder in Claude Code — either the Claude Desktop app or the terminal (cd ~/Documents/GitHub/[your-business] && claude). Claude reads the files in that folder; the Main Branch plugin makes /mb-start and the other skills available on both surfaces. |
Why can't Claude find /mb-start (or any /mb- command)? | Main Branch ships as a Claude Code plugin. Install it once with claude plugin marketplace add noontide-co/mainbranch, then enable it. The plugin works in both the Claude Desktop app and the terminal claude CLI, and it survives git worktrees. Cloud/web Claude sessions can't load plugins — use Desktop or the terminal. If it is installed but skills still don't show, run mb status --json --peek and mb doctor repair --plan from the business repo, apply the approved repair, then restart Claude. |
I see .mb/, not .mb-vip/. Is that wrong? | You're good. .mb/ is the current repo-local Main Branch state folder. .mb-vip/ was old clone-based setup language and is not required. If slash commands are missing, use /mb-update or run mb update --repo ., then repair skill links and restart Claude. |
| Do I start Claude in the business folder or site folder? | Start in the business folder for strategy, bets, decisions, offers, and routing. Switch to a linked site folder only for site implementation work. Linked site folders should have .mainbranch/repo.json; older site folders may still use .mainbranch/source.json. |
| When use skill prompts? | For structured tasks: /mb-start, /mb-think, /mb-ads, /mb-site, /mb-organic |
| Drag files in? | Drag from Finder into Terminal, path appears |
| Voice input? | Wispr Flow (affiliate link) |
| What is content-strategy.md? | The business-level content strategy and index: what you want to be known for, who content is for, pillars, content jobs, and what not to publish. Use core/marketing/ and core/people/ only when you need channel, account, or person-specific layers. |
| How do I build a content strategy? | Run /mb-think. Start with offer clarity, recognition target, audience, and 3-5 pillars in core/content-strategy.md. Add distribution, channel, account, and person files only when the simple file gets crowded. |
| How do pillars work? | Each pillar is a content theme that passes three tests: Soul test (connects to why), Offer test (leads to mechanism), Audience test (they care). 3-5 pillars cover your content universe. |
| What's the content pipeline? | Strategy first, then distribution. Owned assets, email, social, communities, wiki, changelog, and paid amplification should point at the same offer and recognition target. Specific execution lives in pushes/; results land in log/. |
| What are subagents? | Claude can spawn parallel agents to research or review simultaneously. You'll see it happen automatically in /mb-think (multi-source research) and /mb-ads review (6 compliance lenses). Each agent gets its own context window so your main conversation stays clean. |
| How do I manage context/tokens? | Context management is a skill that develops over time. Your files (core/, research/, decisions/) survive compaction — only conversation memory compresses. After compaction, help Claude rebuild by pointing it at recent files or running /mb-start. Save insights to research files early — if it's in a file, it's safe. |
| How do I just save my progress mid-session? | Say "save" or "checkpoint this" — the agent runs mb checkpoint, shows the planned save, and saves on your approval, then you keep working. No need to close the session. Saves always go through mb checkpoint, never raw git. |
| How do I close a session? | Run /mb-end. It summarizes what happened, asks if you have final thoughts, offers a crystallize moment if you made decisions, and guides an approved checkpoint instead of raw git commits. Bookend to /mb-start. A quick mid-work save is just mb checkpoint (above) — /mb-end is the full close. |
| What is multi-offer? | Multiple products under one brand, one business folder. Each offer gets its own core/offers/[name]/offer.md. Soul and voice stay in core/ because they're brand-level. Use when you sell multiple things (community + newsletter + done-for-you). If you have no core/offers/ folder, you're in single-offer mode — everything reads from core/ and nothing changes. |
| How do I switch offers? | Say /mb-start [offer-name] or answer when /mb-start prompts. The choice is session-scoped unless you explicitly approve saving it as local active-offer state. |
| Where do offer files go? | In a single-offer repo, core/offer.md is the durable offer truth. In a multi-offer repo, core/offer.md is the portfolio thesis and core/offers/[name]/offer.md holds offer-specific details. Use current paths only; if CWD looks like an old Main Branch repo, run the repair/migration plan before relying on it. |
| What stays in core with multiple offers? | soul.md (always), voice.md (always), audience.md (base, with optional per-offer overrides), content-strategy.md (business-level content strategy), and optional core/marketing/ / core/people/ layers. |
| Where does proof go? | Company-wide testimonials go in core/proof/testimonials.md, average-case context in core/proof/typicality.md, and angles in core/proof/angles/. Offer-specific proof uses matching files under core/offers/[name]/proof/. Use structured permission and offer-link fields when proof should be detectable by mb status. |
| Can I delete or rename an old offer folder? | Ask first. Offer folders are durable business history. Rename, delete, merge, or move them only after an accepted decision, approved migration plan, or explicit operator instruction. |
| How do I add another offer? | Run /mb-setup -- it detects your existing setup and offers a migration path. Or use /mb-think to plan the new offer first. |
| Do I need separate folders for separate businesses? | If they share soul.md and voice.md, use one business folder with offers/. If they have different identities, use separate folders. The test: shared soul = shared folder. |
| Do my files disappear when context compacts? | No. Compaction compresses Claude's conversation memory, not your files. Everything in core/, research/, and decisions/ is on your hard drive, version controlled with git. If it's in a file, it's permanent. |
| How do I recover after compaction? | Point Claude at recent files: "look at my last 3 decisions" or "read the commits from today." Or just run /mb-start — it scans your folders and rebuilds context automatically. You can also open files yourself in Cursor, Warp, VS Code, or any text editor. |
| Is this system finished? | Still new and actively being tuned. We're building around Claude Code, minimizing commands you need to learn while giving you real power. There's progressive discovery in /mb-think — the more you use it, the more it reveals. You might find workflows we haven't documented yet. Post them in Skool. |
| Where can I see my files? | Open your business repo folder in Finder, Cursor, Warp, VS Code, or any text editor. They're regular .md files on your hard drive. GitHub Desktop also shows them with version history. |
| Where do my customers/leads live? Can my agent see them? | Declare it once: mb spine declare --store <provider> records which system holds your people as a repo fact; mb doctor then grades it honestly. Full map of newer surfaces: cli-surfaces.md. |
| How do I know my money path still works overnight? | mb canary init scaffolds a smoke harness + alert doctrine the moment real money or leads flow unattended. See cli-surfaces.md. |
| How do I get a daily read on the business? | mb pulse init scaffolds deterministic collectors + a repo-local pulse skill that writes one short daily paper with exactly one recommended action; mb pulse install emits the daily run wrapper + cron line (you activate it). See cli-surfaces.md. |
| We're taking real money — is my repo protected? | mb production plan reports the branch-protection gap (PR required, CI + canary gates, force-push/deletion blocked) and prints the exact gh apply commands; applying is yours to run. See cli-surfaces.md. |
| This ad set is cheapest per lead — but are the leads real? | mb leads grade --file leads.json --spend N grades each lead (email, URL, owner answer) and shows ELIGIBLE-lead CPL beside raw CPL, so budget calls use the honest number. See cli-surfaces.md. |
| Which ad creatives are working — keep or kill? | mb ledger init scaffolds the creative ledger (per-asset spend, raw vs eligible CPL, downstream event, KEEP/KILL/WATCH) — an asset isn't working until it makes an eligible lead. See cli-surfaces.md. |
| Is my site ready to launch? | Use /mb-site; the agent runs mb launch check for app/deploy/commerce/email/smoke posture and mb site check for paid-traffic measurement detail. Publishing, spend, provider changes, and customer contact still need your explicit approval. |
| I run an overnight loop but can't see its state. | mb automation init scaffolds the steered-loop contract (loop-state.md): one inspectable file the loop reads first and updates each run; handoffs render from it. See cli-surfaces.md. |
| How do scripts read credentials safely? | mb connect token <provider> prints the credential to stdout for the consumer only — built for scheduled tasks and agents, never for chat. Custom tools: mb connect <id> --custom. |
| Are any of my keys sitting in plaintext? | mb connect hygiene scans agent configs (~/.claude.json, project MCP/settings) for plaintext credentials and shows where to move each one (keychain/env). Read-only; never prints the value. |
| What's my ad account / pixel / page id? | mb connect identity reads the recorded identity facts across connected providers and names what's unrecorded — agents read identity from facts, never infer it from live state. |
| What should I connect first? | Run mb connect plan from your business repo. It shows GitHub, Cloudflare, Google/Workspace, Meta Ads, and Apify as numbered choices with readiness and exact next commands. |
| Skills not showing in a workspace-isolated tool? | Run mb skill link --repo . from the business repo when possible, then restart Claude. Some tools also support a pre-start script that creates bridge links and settings.local.json; see workspace-setup.md for the generic pattern. |
| What's a pre-start script? | A workspace tool hook that runs before the agent starts. Use it to set up Main Branch bridge links only when mb skill link --repo . is not enough. |
Create and review ads, and prepare provider-safe launch plans/checks. Flexible entry points: full pipeline (copy + images), copy only, images only, creative variations (hook library), video scripts, long-form video ads/VSL-style paid creative, video repurpose, compliance review, launch-plan, instrumentation, check, or optional Meta ad account check. Use when asked to create ads, ad copy, image prompts, video ad scripts, paid sales videos, creative variations, review ads, plan paid traffic, check analytics/GA4/GTM/pixels/booking/form readiness, or check launch performance. Say /mb-ads or describe what you need.
Combined research, decision, and codification workflow. Use when: (1) Exploring a question before committing (2) Making a decision that needs documentation (3) User says research, decide, figure out, explore, codify, enrich, retire stale source, sharpen this offer, improve the offer, add context, keyword gate, analyze this VSL, extract the pitch, script teardown, objection mining, or sales video research (4) Updating reference files based on decisions (5) User wants to add testimonials, angles, proof, mechanisms, objections, guarantees, or pitch context. Supports modes: full flow, research-only, keyword-gate, decide-only, codify.
Triage and build any site shape -- lander (1 page), minisite (~4 pages), or full website -- and graduate between them. Also writes owned-surface sales videos/VSLs, about-page videos, landing-page videos, and embedded pitch scripts. Routes to per-shape build flow, reads from business context files, deploys to Cloudflare Pages with git auto-deploy. Use when: (1) Operator says 'I want a site' / 'I want a lander' / 'spin up a one-pager' (2) Setting up a new site of any shape from offer + audience material (3) Updating / iterating on an existing site (4) Graduating a site to a new shape (lander -> minisite -> website -> website + CMS) (5) Writing a sales video, VSL, about-page video, landing-page video, or embedded pitch script for an owned conversion surface (6) Checking analytics/instrumentation, GA4, GTM/dataLayer, Meta pixel, booking links, HubSpot/forms, conversion-event mapping, form-submit smoke, or paid-traffic readiness (7) Previewing or publishing changes Triggered by: /mb-site, 'build a site', 'landing
Update Main Branch. Use when the user says update, upgrade, pull latest, get latest Main Branch, asks whether they are current, or after Devon announces a new release.
Bootstrap a new business folder with Main Branch structure, or migrate an existing single-offer folder to multi-offer. Use when: (1) New user needs Claude Code environment configured (2) User says "set up", "get started", "initialize", "bootstrap", "create my repo", "new business" (3) User is new to Main Branch and needs full onboarding (4) Migrating existing business context into the Main Branch structure (5) User wants to add a second offer to an existing repo. Creates CLI and skill wiring, a business folder with full structure, and bounded context until core files are useful. Applies business setup patterns. Teaches concepts during setup.
Pre-publish review / UI pass / design pass / design critique for a brief or copy draft: runs dial-gated Seven Sweeps + auxiliary gates and returns synthesized findings. Use when: the operator asks for a pre-publish review, a UI pass, a design pass, a design critique, or to review a page before publishing. Called by /mb-site at pre-lock and pre-publish; reusable by /mb-ads and sales-video workflows.