Product support

You need to be in a conversation inside your Project, not a loose chat. Create one from Projects in the left sidebar, open it, start a new conversation there, and say set up my Accountant OS again (or type /accountant-setup-wizard). The Project is where your profile and trackers are saved.

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Work inside the Project where you ran setup. If you skipped setup, say set up my Accountant OS — paste your website or an email you've sent a client and it fills in most of your profile in two minutes. Say set up my Accountant OS again anytime to update it.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output and save files in your Project.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Tax Client File — Patel household." Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Scheduled runs prepare everything and save it for your review — they don't write to your email, calendar, or Drive on their own. Open the conversation it created, review, and say "move this week's follow-ups to Gmail."

Each keeps its own setup, connections, and trackers, so both can live in one Project. For this one, type /accountant-setup-wizard and /accountant-connect-tools rather than the shared phrases — then run the Bookkeeper product's own setup and connect steps the same way. Its jobs and trackers have different names, so they don't mix.

Say "open my command center" from anywhere — it drops what was in progress and shows home again.

This is the failure mode to watch most closely. Two fixes:

• Verify every IRC section, Reg, Rev. Proc., and dollar threshold against current authority before anything goes to a client. tax-specificity-guard flags drafts that state a citation precisely so you don't miss one.
• When you catch a fabricated cite, paste it back and ask: "Where did you get this citation? I didn't provide it." Then add to your request: "Never cite authority I haven't provided — use [verify citation] as a placeholder instead."

Say set up my Accountant OS → Complete my profile and list those exact phrases under "words I never use" — every skill reads them. If one keeps slipping through, add a hard line to your request: "Do not use the word 'write-off' in any client-facing document."

Add to your request: "Include 'as of [TAX YEAR]' on any reference to current tax law, rates, or thresholds." A planning memo without it becomes a problem when the client reuses it next year.

• Your Project folder is probably thin. Add 3–5 of your best anonymized client letters so Claude can see your actual voice, then say set up my Accountant OS → Complete my profile to sharpen the voice answers.
• Or the request didn't ask for enough specificity. Re-run with: "Match the voice from my sample letters in the Project folder, and reference the specifics of [this client's situation]."

Add an explicit limit: "150 words max." If Claude keeps going over, add: "When I give a word limit, stay under it — cut content rather than exceed it."

That's practice-boundary-guard doing its job — it fires when a request edges into legal advice or work outside the engagement. Re-scope the request to what you're actually engaged to do, or send the client a scope-change note (/scope-change-notification).

You need to be in a conversation inside your Project, not a loose chat. Create one from Projects in the left sidebar, open it, start a new conversation there, and say set up my Bookkeeper OS again (or type /bookkeeper-setup-wizard). The Project is where your profile and trackers are saved.

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Work inside the Project where you ran setup. If you skipped setup, say set up my Bookkeeper OS — paste your website or an email you've sent a client and it fills in most of your profile in two minutes. Say set up my Bookkeeper OS again anytime to update it.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output and save files in your Project.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Books File — Acme." Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Scheduled runs prepare everything and save it for your review — they don't write to your email, calendar, or Drive on their own. Open the conversation it created, review, and say "move my close reminders to Gmail."

Each keeps its own setup, connections, and trackers, so both can live in one Project. For this one, type /bookkeeper-setup-wizard and /bookkeeper-connect-tools rather than the shared phrases — then run the Accountant product's own setup and connect steps the same way. Its jobs and trackers have different names, so they don't mix.

Say "open my command center" from anywhere — it drops what was in progress and shows home again.

You don't answer it — you route it. Run /tax-or-bookkeeping-triage (acknowledge and send to their CPA) or /refuse-tax-advice (a friendly decline). This is the most important habit in the whole product; the cpa-boundary-guard is there to catch it if you forget.

No, that's cpa-boundary-guard doing its job. It fires when a request drifts into tax advice, worker classification, payroll-tax, sales-tax/nexus, audit language, or entity choice. Re-scope to the bookkeeping side (record the facts, route the determination), and you're good.

client-funds-control-guard reframes "what to pay" into "what the client should review and approve." You prepare the list; the client or an authorized approver decides. Stay out of custody risk.

• Your Project folder is probably thin. Add 3–5 of your best anonymized client emails, then say set up my Bookkeeper OS → Complete my profile to sharpen the voice answers.
• Or the request didn't ask for specificity. Re-run with: "Match the voice from my sample emails in the Project folder, for [this client's situation]."

That's exactly what tax-specificity-guard flags — verify it before sending. Never let a stated number go to a client unverified; use [verify current rule] until you've confirmed it.

Add an explicit limit: "150 words max." If it keeps going over, add: "When I give a word limit, stay under it — cut content rather than exceed it."

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Say set up my Real Estate OS again. It offers to update your profile, complete it (the full setup adds voice, market, and compliance detail for sharper output), start fresh, or go to your command center.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output. Also note: only the five Into your tools jobs write to Gmail, Calendar, and Drive. The weekly-rhythm jobs (listings, deals, daily outreach, Friday loop, listing appointment, buyer) build a board and save files in your Project by design.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job its name: "use Seller Updates — 642 19th Ave N" or "update Transaction Board — 12 Oak St." Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Say my listings to see what's saved. Add or fix them there, or import a CSV or Excel export: attach it with the + in the message bar (or drop it into your Project folder) and say import it.

Scheduled runs only prepare — they build the board and save the drafts for your review, and never send, post, or write to your tools on their own. Open the conversation it created and review.

Say "start over" or "open my command center" from anywhere — it drops what was in progress and shows home again.

That's the one thing to watch. The rule every job follows: if you didn't supply it, it becomes a [confirm] placeholder, not a guess. There's no live MLS access — pull the comps yourself and paste them in. If you see a number or feature you didn't give, replace it and tell the job to use only what you provided.

The fair-housing check runs automatically on every job and prompt headed for a client, the MLS, or the public, and ends with a before-publishing checklist. Still do a final read with one rule in mind: describe the property and the location, never the buyer. Marketing a 55+ community? Age-targeted language is only allowed with a "confirm valid 55+ status for this listing" stamp — confirm it from the actual listing.

The jobs won't answer those, by design — the practice-boundary guard reframes them to agent-supplied, professional-reviewed. Draft the part that's yours and route the rest to the attorney, CPA, or lender.

Deadlines are never calculated for you — contracts vary. Give the job the date from the contract ("add the appraisal deadline, June 20, to 12 Oak St") and it updates the board (and your calendar, if connected).

Stop. Nothing here originates, relays, or confirms wire instructions, and neither should an email it drafts. Tell the client to use the verified instructions they already received from title or escrow and to confirm them by phone using a number they independently know before sending a cent.

You need to be in a conversation inside your Project, not a loose chat. Create one from Projects in the left sidebar, open it, start a new conversation there, and say set up my Loan Officer OS again (or type /lo-us-setup-wizard). The Project is where your profile and trackers are saved.

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Work inside the Project where you ran setup. If you skipped setup, say set up my Loan Officer OS — paste your website, your email signature, or an email you've sent a realtor partner, and it fills in most of your profile in two minutes. Say set up my Loan Officer OS again anytime to update it or complete the full profile.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output and save files in your Project. If your lender doesn't allow borrower work in a personal Gmail or Drive, leave them unconnected — paste-ready is the right mode for you.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Loan Tracker — Garcia." Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Scheduled runs prepare everything and save it in your Project for review — they don't write to your email, calendar, or Drive on their own. Open the conversation it created, review, and say "move my Friday updates to Gmail."

Jobs only put dates on your calendar that you gave them — an expiration date, a lock expiration, a contract or closing date. They never compute a date. Give the date and run the job again.

Say "open my command center" from anywhere — it drops what was in progress and shows home again.

It shouldn't, and the design prevents it — but if a number appears, never send it. Rates, APR, payments, points, and lock terms always come out as [from engine] / [from LE] placeholders you fill from your pricing engine and the Loan Estimate. If a number slips in, replace it with a placeholder and add to your request: "Never state a rate, APR, payment, or lock term — use [from engine] placeholders only." lo-compliance-guard flags drafts that drift toward a quote.

That's a UDAAP trap — stop. Add to your request: "No future-rate predictions, no urgency or scarcity claims, no guaranteed-savings language." The guard flags these, but the judgment is yours.

The pack drafts the copy; RESPA Section 8 compliance is yours. Confirm the cost split reflects fair market value and you've documented who paid for what before publishing. If you're unsure whether a piece crosses the line, send it to your compliance officer first — Section 8 violations carry per-occurrence penalties.

Never describe a borrower by a protected class. Add to your request: "Describe the loan, the property, and the math — never the borrower's protected characteristics." Re-run and scrub before sending.

• Your Project folder is probably thin. Add 2–3 of your best real borrower or realtor emails (scrubbed of names and NPI) so Claude can see your actual voice, then say set up my Loan Officer OS → Complete my profile to sharpen the voice answers.
• Or the request didn't ask for enough specificity. Re-run with: "Match the voice from my sample emails in the Project folder, and reference the specifics of [this borrower's situation]."

Add an explicit limit: "100 words max." If Claude keeps going over, add: "When I give a word limit, stay under it — cut content rather than exceed it."

Use /denial-with-options for the borrower communication, but the formal adverse-action notice still goes through your compliant process — it has ECOA timing and content rules the guard flags. The draft is the message, not the legal notice.

/lo-us-appraisal-update communicates appraisal status (ordered / received / delayed) and states no value. It does not negotiate value, advise on a reconsideration-of-value (ROV), or counsel the contract — that's a live conversation with the borrower and agent, not an AI draft.

You need to be in a conversation inside your Project, not a loose chat. Create one from Projects in the left sidebar, open it, start a new conversation there, and say set up my Contractor OS again (or type /contractor-setup-wizard). The Project is where your profile and trackers are saved.

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Work inside the Project where you ran setup. If you skipped setup, say set up my Contractor OS — paste your website or an email you've sent a customer and it fills in most of your profile in two minutes.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output and save files in your Project.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Job Tracker — Miller Kitchen." Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Scheduled runs prepare everything and save it for your review — they don't write to your email or calendar on their own. Open the conversation it created, review, and say "move these to Gmail."

Say "open my command center" from anywhere — it drops what was in progress and shows home again.

This is the failure mode to watch most closely. Two fixes:

• Verify every code citation, permit requirement, and inspection rule against your local building department or your reference materials before anything goes to a customer or the city. Codes are local and they change.
• When you catch a fabricated reference, paste it back and ask: "Where did you get this code citation? I didn't provide it." Then add to your request: "Never cite a building-code section I haven't provided — use [verify with local code] as a placeholder instead."

Always check the math yourself — Claude is not a calculator and doesn't know your market. Give it your actual numbers in the prompt; don't ask it to estimate prices for you. On change orders and estimates it's built to mark [verify total] rather than calculate — keep it that way.

That's the warning sign to stop and route it to your attorney. The pack is explicit that contract, lien, and waiver instruments come from counsel — lien-and-license-guard flags these and appends a checklist, but the actual legal step is the attorney's, and a statutory deadline is yours to track. Never let a lien or notice window slip.

Add to your request: "Use 'estimated,' 'approximately,' and 'subject to site conditions.' Never write 'guaranteed' about price, timeline, or outcome." The advertising-environmental-scope-guard also flags guarantee language and "licensed/insured/bonded" claims for substantiation.

• Your Project folder is probably thin. Add 2–3 of your best real customer emails (scrubbed of names and numbers) so Claude can see your actual voice, then say set up my Contractor OS → Complete my profile to sharpen the voice answers.
• Add the phrases you'd never say to the "words I never use" list, and the phrases that sound like you to the voice samples.

Add them to your "words I never use" list, and add a hard line to the request: "Use plain language. No fancy construction or corporate jargon."

• Your Project context is probably empty. Add 2–3 real estimates and emails (scrubbed).
• Your trade and specialty in the setup are probably too broad. Get specific.
• Add specificity to the request: "This is for a [specific job type] customer in [specific situation]."

For an everyday field change to scope, cost, or schedule, use /change-order. Only use /contract-addendum-cover-letter when you're routing a formal, attorney-prepared amendment to the contract. When in doubt, it's /change-order.

That's advertising-environmental-scope-guard or lien-and-license-guard doing its job — they fire when a request edges into legal advice, engineering, specialty-license work, or another trade. Re-scope the request to what you're licensed to do, and route the rest to the right professional.

You need to be in a conversation inside your Project, not a loose chat. Create one from Projects in the left sidebar, open it, start a new conversation there, and say set up my Photographer OS again (or type /photographer-setup-wizard). The Project is where your profile and trackers are saved.

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Work inside the Project where you ran setup. If you skipped setup, say set up my Photographer OS — paste your website or an email you've sent a client and it fills in most of your profile in two minutes. Say it again anytime to update your profile or add your AI-edit stance.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output and save files in your Project.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Client Tracker — Parker Wedding" (or "use the Lead Board"). Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Scheduled runs prepare everything and save it for your review — they don't write to your email or calendar on their own. Open the conversation it created, review, and say "move these to Gmail."

Say "open my command center" from anywhere — it drops what was in progress and shows home again.

• Say set up my Photographer OS → Complete my profile and add them to your "words I never use" list explicitly. Every skill reads it, and photographer-voice-keeper enforces it. Common bans: "capturing memories," "freeze the moment," "tell your story," "magical," "epic," "obsessed," "dream session," "your forever."
• At the end of a prompt you can add: "Don't use any of the clichés from my voice list."

• Your Project folder is probably thin. Add 3-5 of your best real inquiry replies, a gallery delivery email, and a few captions (scrubbed of names) so Claude can see your actual voice, then say set up my Photographer OS → Complete my profile to sharpen the voice answers.
• Add to the prompt: "This must sound specifically like me, not a generic photographer. If any studio could have sent it, rewrite."

• Upload more of your best converting replies — voice comes from your samples, not defaults.
• Always provide the specific detail from their inquiry (venue, date, what they asked) so the reply can be specific to them.

• Add an explicit limit: "100 words max." If Claude keeps going over, add: "When I give a word limit, stay under it — cut content rather than exceed it."

It shouldn't — the design prevents it. If a number appears that you didn't provide, never send it. Replace it with a placeholder and add: "Never state a price, date, or deliverable count I didn't provide — use a placeholder." When you catch one, ask "where did you get this number?" and reinforce the rule.

Add: "All contract-clause output must begin with 'DRAFT — REVIEW WITH ATTORNEY IN [STATE] BEFORE SENDING' and end with a list of points to flag." Have one attorney review all your clauses once — especially the AI-edit, AI-training, and usage-rights language.

No — that's fabrication and breaks platform rules. /review-testimonial-request asks the client for an honest review and makes it easy. If a client shared real words you want to display, the pack tidies their words but stamps it a draft for them to approve and post themselves — never a quotable testimonial it invented.

/reschedule-cancellation-reply drafts the human message. It points to your contract's reschedule/retainer/cancellation terms as you supply them — it never invents a policy. When money is in dispute, review against the signed agreement before sending.

/final-balance-reminder drafts the courtesy reminder. The amount and due date come from your signed contract as placeholders — if you don't supply them, the draft shows placeholders rather than a guess.

Vendor pitches are about a specific shared relationship. Always provide the specific shared client, venue, or context in the prompt — a pitch with no specific shared context will always sound generic. Use /vendor-pitch-wedding or /vendor-pitch-brand.

You need to be in a conversation inside your Project, not a loose chat. Create one from Projects in the left sidebar, open it, start a new conversation there, and say set up my Social Media Manager OS again (or type /smm-setup-wizard). The Project is where your profile, client cartridges, and trackers are saved.

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Work inside the Project where you ran setup. If you skipped setup, say set up my Social Media Manager OS — paste your website or an email you've sent a client and it fills in most of your profile in two minutes. Say it again anytime to update your profile or add a client.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output and save files in your Project.

That's by design — no job posts or schedules anything. Every post comes out ready to paste into your own scheduler, and each graphic comes as a design note to build in Canva.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Content Tracker — [client name]." Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Run "onboard a client" for them (or paste 3–5 of their captions) — jobs won't guess a client's voice.

Scheduled runs prepare everything and save it in your Project for review — they don't write to your email, calendar, or Drive on their own. A scheduled report can't pull platform numbers either: it prepares the report shell and a "numbers to export" checklist. Open the conversation it created, review, and carry on from there.

Say "open my command center" from anywhere — it drops what was in progress and shows home again.

Three likely causes:

• The cartridge isn't specific enough yet → re-run /cartridge-builder after a couple weeks of real posts.
• You didn't tell Claude which client this is, so it didn't pull the right cartridge → name the client in your request.
• Add "never use corporate jargon" to that client's cartridge never-use list as a blanket rule.

• The cartridge is too thin → re-run /cartridge-builder to refine it, and add 3–5 real sample posts from this client so Claude can see the actual voice.
• Or the prompt didn't ask for enough specificity → re-run with: "This is [Client]. Match the style of their cartridge's sample posts closely."

TikTok wants a different modulation — more casual, faster. Add platform-specific notes to the cartridge describing how the voice shifts per platform, or re-run /caption-tiktok with "make this more casual / faster-paced than the Instagram version."

This is the failure mode that loses client confidence fastest — prevent it obsessively:

• Work one client per request, and never ask for two clients' posts in the same request.
• Name the client so Claude pulls only that cartridge.
• Run /voice-check to compare a draft against the right cartridge, and use the 60-second voice-bleed scan in cheat-sheet.pdf before anything from a different client goes out.

After the first run you should be editing for under two minutes, not ten. If it's more, the cartridge isn't aligned with reality (re-run /cartridge-builder) or the AI is defaulting to generic language (add more real sample posts to the cartridge). Tighten the feedback loop — see why-this-works.pdf.

Your /content-calendar context didn't specify distribution. Re-run with: "Mix pillars throughout — no more than 2 consecutive posts in the same pillar." If the imbalance persists, review the cartridge's pillars with /cartridge-builder.

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output and save files in your Project.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Engagement Tracker — Northwind." Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Scheduled runs prepare everything and save it for your review — they don't write to your email, calendar, or Drive on their own. Open the conversation it created, review, and say "move my Friday status to Gmail."

Say "open my command center" (or "start over") from anywhere — it drops what was in progress and shows home again.

The pack is built to use real, attributable sources and to flag uncertain ones rather than invent them — where a specific statute or CFR section is uncertain you'll see "[verify exact citation before client use]". Treat that as a feature: verify any specific citation before it goes to a client. If you ever see a confident-looking citation you can't source, don't ship it — ask the skill where it came from.

The pack won't answer those, by design — practice-boundary-guard reframes anything that strays into a legal/tax/audit/valuation/regulatory determination to a recommendation for the client's licensed advisor. Draft the part that's yours (the strategy, the options, the questions for counsel) and route the determination to the right professional. That's not a limitation; it's what keeps you inside your scope of practice.

Use /linkedin-post-from-engagement — it's built NDA-safe (no client names, no deal specifics), and confidentiality-guard double-checks it. The rule it enforces: share the lesson, never the client. If you paste client-identifying details, the guard will flag them with a redaction checklist before anything is public-ready.

If there's any chance of a conflict, run /conflict-of-interest-check before you accept — it walks a structured COI check so you catch the problem before you're committed.

You need to be in a conversation inside your Project, not a loose chat. Create one from Projects in the left sidebar, open it, start a new conversation there, and say set up my Financial Advisor OS again (or type /financial-advisor-setup-wizard). The Project is where your profile and trackers are saved.

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude. (Anthropic's own Claude for Financial Services plugin is different — it's designed to sit alongside this one, so keep it if you use it.)

Work inside the Project where you ran setup. If you skipped setup, say set up my Financial Advisor OS — paste your website or your email signature and it fills in most of your profile in two minutes. Say set up my Financial Advisor OS again anytime to re-run it or complete your profile.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output and save files in your Project. If your firm requires client email through an approved, archived system, leave Gmail unconnected and paste the drafts there.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Client File — Household Lee." Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Scheduled runs prepare only — they save drafts and files in your Project and update the trackers, but they don't write to your email, calendar, or Drive on their own. Open the conversation it created, review, and say "move these to Gmail."

Say "open my command center" from anywhere — it drops what was in progress and shows home again.

Stop and remove it. The jobs are built to use only the figures you supply and to mark gaps [ADVISOR TO CONFIRM]. Paste the number back and say: "Where did this come from? Use only the figures I provided." Never let a return, projection, or computed deadline reach a client.

Add to your request: "Describe process, not outcomes. No performance claims, projections, or guarantees." The compliance-boundary guard flags this language; your CCO has the final say.

The jobs record only recommendations you state. If a brief or email implies one, delete it and tell the job exactly what you recommended — or that you made none.

That's the boundary: wills, trusts, and beneficiary effect go to the client's attorney; filing positions go to the CPA; insurance products go to an appropriately licensed partner. The jobs record the topic as discussed and route it — they don't answer it.

Meeting follow-up & notes stops on a complaint: it keeps it out of the client email and drafts an internal routing note for your CCO. Follow your firm's complaint procedure.

Don't paste SSNs, full account numbers, or balances tied to a name in the first place — use a client reference and last-4 masks. The privacy guard flags raw NPI, but the redaction habit is the real protection. Scrub any output before it goes anywhere.

You need to be in a conversation inside your Project, not a loose chat. Create one from Projects in the left sidebar, open it, start a new conversation there, and say set up my Founder OS again (or type /operator-setup-wizard). The Project is where your profile and trackers are saved.

Your jobs still find your earlier setup. To move it into your Project (recommended — the jobs can then save what they learn), say set up my Founder OS and paste your old profile block; it carries every answer over without re-asking.

Work inside the Project where you ran setup. If you skipped setup, say set up my Founder OS — paste your website or an email you've written to a customer and it fills in most of your profile in two minutes.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output and save files in your Project.

Those connectors are optional and read-only. Say connect my tools to see which ones this session can use. Without them, paste an export or save it in your Project — every job works from pasted data.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Founder Metrics tracker" (or whichever it is). Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Scheduled runs prepare everything and save it for your review — they don't read your payment tools or write to your email or calendar on their own. Open the conversation it created, review, and say "move these to Gmail" or "add my focus blocks."

Say "open my command center" from anywhere — it drops what was in progress and shows home again.

It shouldn't be — every job marks a figure it doesn't have as [confirm] instead of estimating. If you catch one, ask "where did this number come from?" and give it the real figure. Check the math in any workbook yourself; runway scenarios are spreadsheet formulas so you can.

Guards are not vetoes. Read the checklist, decide whether the concern applies, and carry on. They raise a question; the decision is yours.

Hiring, firing, contractor classification, and anything that reads as a fundraising offer go to a professional. The system flags these — it doesn't decide them.

• Add a one-page product description and 3–5 customer emails you're proud of (with names removed) to the Project folder.
• Say set up my Founder OS → Complete my profile and paste better voice samples.
• Add to the request: "My product is unusual in that [X]. Keep that in mind."

Output will be generic — it can't see your profile or trackers. Move to a conversation inside your Project and run it again.

Open Customize → Plugins → Yours and check it's listed with its toggle on. If it isn't, upload it again: Customize → Plugins → Add ▾ → Upload plugin, and pick the plugin zip as-is — don't unzip it. Then open a conversation inside your Project and type / again. If you got an error about nested zips, you uploaded the documentation zip — upload the smaller plugin zip instead.

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Say connect my tools. It checks each connection, tells you what's missing (connectors are added under Customize → Connectors), and asks whose inbox and calendar it's working in — your executive's through delegated access, your own, or both. Until then, jobs hand you paste-ready output and save files in your Project. Whether a connector is allowed on your executive's accounts is your company's call — when in doubt, check with IT first.

That's by design — nothing touches the inbox, the calendar, or Drive without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job which one: "use the Trip Tracker — Denver · Oct 20–22" or "use the Scheduling Board." Anything named "SAMPLE — …" came from the sample run and is ignored on purpose. If your waiting-on list or action register lives in another tool, say so in your profile — the jobs will give you paste-ready rows for it instead of keeping a second list.

Scheduled runs prepare only: they write to your trackers and save the drafts as files in your Project, and never touch Gmail, the calendar, or Drive on their own. Open the conversation it created, review, then say "move today's replies to Gmail" (or "move this week's plan to Gmail and place the holds").

Say "open my command center" from anywhere — it drops what was in progress and shows home again.

That information lives in the task, not the profile — so it doesn't carry across tasks. Anything you want every skill to know (your executive's working preferences, your systems, your approval authority, your standing rules, both voices) belongs in the profile: say set up my EA OS → Complete my profile and add it there once. Per-task details (today's calendar, this trip's confirmations, what came out of this morning's meeting) you provide each time — that's by design, since they change daily.

Voice comes from real samples, not adjectives. Say set up my EA OS → Complete my profile and paste 2-3 pieces of actual writing for each voice, with third-party names stripped: for yours, a scheduling note or decline you were happy with; for your executive's, messages they wrote themselves that are unmistakably them. The wizard's voice batch is the highest-value part of setup and the one people rush.

If the content rather than the tone is generic, the profile is thin rather than the samples. A day plan written without knowing your executive's meeting rhythm, focus-time habits, or what they always want handled without being asked is a day plan that could belong to any executive in the country. Tighten those in your profile and the difference is immediate.

Re-run /ea-command-center. The fixed layout is the contract — today, your waiting-on list, open follow-ups, and the menu, in that order, every time — and consistency is deliberately the product here. You should be able to find the thing you're looking for by muscle memory rather than reading.

If it drifted, it's almost always because you asked for something mid-task that reshaped the output ("can you also show me next week," "just give me the follow-ups"). That's fine as a one-off, but it changes that rendering, not the layout. Run the command fresh and you're back to the standard view. If sections come back thin, that's not drift — it's that the waiting-on list and register don't have much in them yet. Run a job or /end-of-day-loop for a few days and they fill. One change is expected: the Try it on a sample card disappears for good once your first real tracker exists.

That's executive-confidentiality-guard working, and it's working on purpose. An EA sits inside the most sensitive information stream in the company — personnel moves, compensation, deals, legal matters — and third-party names in those contexts don't belong in an AI tool unless your company has approved that tool for them. Most haven't, for most tools.

So the skills are built to work from role labels: "the candidate for the VP role," "the CFO," "Director K." The guard swaps what it finds, tells you exactly what it swapped, and you re-add the name yourself in the approved system — the ATS, the HRIS, your email — after you've reviewed the draft. That's where the information is supposed to live anyway. The operating rule: follow your company's data policy; when in doubt, keep it in approved systems.

Your executive's own name and context are a different case, and the guard treats them differently — that's what the profile is for. What the guard cares about is what leaves.

Also by design — that's professional-boundaries-guard. Approving spend is a delegated authority that belongs to a named approver under your company's policy, and an AI draft that says "approved" creates a record that isn't true. So /expense-report-narrative will write you the business-purpose narrative line by line, flag the items that need a policy call, and mark the whole thing "prepared for [approver]'s sign-off" — which is the accurate description of what you just produced.

The same guard holds the same line elsewhere: it won't mark anything signed, booked, or sent, it won't commit your executive's time or position without their say ("I'll hold the slot pending his confirmation," not "he'll be there"), and it won't invent a policy threshold — you'll get [verify against your company policy] instead of a number that looks authoritative and isn't. It also won't give legal, tax, HR, or compliance advice, or write a candidate or employee assessment; those route to counsel, finance, HR, and the hiring panel. And anything of value involving a government official gets named as an FCPA flag and routed to legal rather than assessed.

If your authority is genuinely broader than the draft assumes — you can book travel under a threshold without asking — put that in your profile (say set up my EA OS). The guard reads it.

Working as intended. Your executive's lodging, room number, home address, minute-by-minute movements, and family details are physical-security data, not logistics trivia, and executive-confidentiality-guard runs a distribution check on any document that looks like it's going wider than the trip's need-to-know list. The check is one line telling you who the draft is safe to reach as written — "safe for the trip's need-to-know list; NOT for the all-hands channel (hotel and family details)."

You still have the full itinerary. /travel-itinerary produces the complete version for the need-to-know list, and the trimmed version is what you share with the team, the host, or anyone who only needs to know where to be at 2:00. If you need the full detail to reach a specific wider group — a security detail, a spouse, a co-travelling colleague — say so and it'll draft to that audience. The guard flags; it doesn't block.

It knows what you give it. Where your company has approved and enabled a connection to the mail and calendar you use, say connect my tools once and the loop reads them directly — that's the intended experience. Where it hasn't, paste: the day's calendar and the inbox subject lines with sender roles are enough for a good pass.

Whether a connector is available to you is a company decision — approved-tool questions belong to IT and security, and de-identifying inputs doesn't make an unapproved tool approved. The pasted path is fully supported and, for the sensitive half of the inbox, arguably the better one: you control exactly what goes in.

It's built not to. Live logistics come from your records — the skills structure what you give them and write [confirmation number] or [verify with the airline] rather than producing something plausible. Entry requirements in /travel-doc-checklist come back as items to verify with the official source, never as stated fact, because they change without notice and a confidently wrong answer costs a trip. Policy specifics come back as [verify against your company policy].

On frameworks, the pack cites real primary sources — David Allen, Getting Things Done (2001) for the capture-and-review rhythm and the waiting-for list; Paul Graham's "Maker's Schedule, Manager's Schedule" (July 2009) for focus-block defense; Merlin Mann's Inbox Zero (2006-2007) as attention management rather than an empty-inbox target; the Eisenhower matrix credited as popularized by Covey rather than as a system Eisenhower authored; Robert's Rules of Order Newly Revised, 12th edition (2020) only where a body has actually adopted it. Where a practice has no canonical author — timeboxing, the action-register columns, ordinary agenda and minutes convention — it says "general convention" instead of inventing an authority. If you ever see a confident-looking citation, statistic, or policy figure you can't source, don't ship it — ask the skill where it came from.

Yes. The pack ships a raw prompts pack: every prompt in click-to-copy HTML and PDF, plus one plain-text file per prompt. Paste them into ChatGPT, Claude, Gemini, or whatever your company allows — no install, no Project, no plugin. You give up the saved profile, the command center and jobs, the shared state the loops run on, and the always-on guards, so you'll re-state your executive's context each time and the de-identification is entirely on you: use role labels before you paste, and run your own distribution check before anything with travel or personnel detail leaves your drafts folder. Worth knowing: claude.ai on the web is not supported for the plugin path, because the website doesn't run plugins. The Claude desktop app and Claude Code do.

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors). Until then, jobs hand you paste-ready output and save files in your Project. Texts, DMs, and training-app messages are always paste-ready — there's no connector for those.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Client Tracker — Dana." Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Scheduled runs prepare everything and save it for your review — they don't write to your email, calendar, or Drive on their own. Open the conversation it created, review, and say "move my check-ins to Gmail."

Say "open my command center" (or "start over") from anywhere — it drops what was in progress and shows home again.

That information lives in the conversation, not the profile — so it doesn't carry across conversations. Anything you want every skill to know (your certifications, your typical client, your scope rules, your voice) belongs in the profile: say set up my Personal Trainer OS → Complete my profile and add it there once. The five jobs also save missing profile details as you work, with an undo line. Per-client details (this person's screening, goal, equipment) you provide in the conversation each time — that's by design, since they change from client to client.

Voice comes from real samples, not adjectives. Add 2-3 of your actual (de-identified) client messages and a program note to your Project folder, then say set up my Personal Trainer OS → Complete my profile and paste one of them. The voice match improves immediately.

The pack is built to use real, attributable sources — periodization to Matveyev/Bompa and Verkhoshansky/Issurin, progressive overload to DeLorme, RPE to Zourdos (2016), screening to PAR-Q+ and ACSM's current Guidelines — and to use "industry experience consistently shows" rather than invent a statistic. It's also built to not write "muscle confusion," the "anabolic window," lactic-acid soreness, or spot reduction as fact. If you ever see a confident-looking claim or citation you can't source, don't ship it — ask the skill where it came from.

The pack won't cross that line, by design — scope-of-practice-guard reframes anything that strays into diagnosis, injury treatment/rehab, prescriptive nutrition or supplement dosing, or a condition-that-needs-clearance into a clean refer-out. Use /refer-out-conversation: report what the client factually said (no interpretation), name what's needed (physician or PT clearance, a dietitian referral), work around it in the meantime, and align to their clearance once you have it. That's not a limitation; it's what keeps you inside your scope of practice — and out of real liability.

Use /client-transformation-post or /testimonial-request — both are built to share the lesson and not the client, and client-privacy-guard double-checks every public draft. The rule it enforces: no client identity, health info, body metrics, or progress photo goes public without that client's explicit, documented, revocable consent — and the framing stays body-neutral, capability-focused, never before/after shame. If you paste client-identifying details, the guard flags them with a before-you-post checklist.

Same cause, one step earlier: you're in a loose chat rather than a conversation inside your Project, or the plugin didn't finish installing. Open Customize → Plugins → Yours and check the plugin is listed with its toggle on. If it isn't, upload it again: Add ▾ → Upload plugin, and pick the teacher-claude-plugin-…zip as-is — don't unzip it. If you got an error about nested zips, you picked the documentation zip — upload the plugin zip instead. Then open a conversation inside your Project and type / again.

That information lives in the conversation, not the profile — so it doesn't carry across conversations. Anything you want every skill to know (your grade and subjects, your standards framework, your school's period length, your class profiles as counts and levels, your voice) belongs in the profile: say set up my Teacher OS → Complete my profile and add it there once. (Jobs also ask for a missing detail the first time they need it, and save it for you.) Per-lesson details (today's topic, this assignment's rubric, what you saw in the exit tickets) you provide in the conversation each time — that's by design, since they change every day.

Say connect my tools. It checks your district's AI policy from your profile, tests each connection, and tells you what's missing (connectors are added under Customize → Connectors). School accounts sometimes block new apps at the district level — if sign-in is refused, that's your district's IT team's call, and every job keeps working paste-ready with files saved in your Project. Connect only accounts your district allows.

That's on purpose when your families are reached through ParentSquare, ClassDojo, Remind, Seesaw, TalkingPoints, or Google Classroom — none of those have a connector, so the job gives you text sized for the platform to paste in. Email drafts are made only when your family channel is email.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Class Tracker — Period 2." Anything named "SAMPLE — …" came from the sample run and is ignored on purpose.

Scheduled runs prepare only: they draft from your Class Tracker and save the file in your Project for review — they never write to your email, calendar, or Drive on their own, and the scheduled newsletter never drafts individual family emails. Open the conversation it created, review, then say "move my week to Drive" or "move my newsletter to Gmail."

Turn it off in Customize → Plugins → Yours. The Operating System includes everything in it, and two copies of the same prompts confuse Claude.

Say set up my Teacher OS — it offers buttons: Complete my profile (or Update my profile, once everything's filled in), Start fresh, or Open my command center. To get back to home base from anywhere, say open my command center; it drops whatever was in progress.

Two causes. First, the pack needs this class's real detail — a plan written from a topic alone is a plan that could have been written for any classroom in the country. Give it your period length, the standard you're working under, what students already know, and a de-identified class profile ("28 students, 6 multilingual learners, 4 with listed accommodations, this section is right after lunch") and it sharpens immediately. Second, your profile may be thin — say set up my Teacher OS → Complete my profile and tighten your grade, subjects, standards framework, and class context.

Voice comes from real samples, not adjectives. Add 2-3 pieces of your actual writing — a parent email you were happy with, a newsletter blurb, a comment you'd be glad to have read aloud — to your Project folder with student names stripped, then say set up my Teacher OS → Complete my profile and point it at them. The voice match improves immediately.

That's student-privacy-guard working, and it's working on purpose. FERPA protects students' education records, and a teacher shouldn't be pasting names, ID numbers, grades tied to an identifiable student, health information, or family circumstances into an AI tool unless the district has approved that tool under its data-privacy agreements. So every skill here is built to work from de-identified inputs: "Student A," "a 7th grader," roles and observable behavior rather than names. The guard swaps what it finds, tells you what it swapped, and appends a before-you-send privacy check. You re-add the name yourself in your district email or gradebook after you've reviewed the draft — which is where student information is supposed to live anyway. The operating rule is simple: follow your district's data policy; when in doubt, de-identify.

Also by design — that's professional-boundaries-guard. Under IDEA, an IEP is written by the IEP team, and a 504 plan by the 504 team. A classroom teacher implements the accommodations and modifications those teams list; they don't write, alter, or second-guess the plan on their own, and no AI draft should make it look like they did. What the pack will do is /accommodation-planner: you give it the accommodations the team already produced, and it turns them into concrete implementation moves for a specific lesson. Anything touching goals, services, eligibility, or placement gets routed to the case manager and the team. The same guard blocks diagnosing or labeling a student, compliance guarantees, invented district policy, and acting on an AI-detector score — each one a place where a confident-sounding draft could genuinely hurt you.

Stop, and use your school's protocol — not this pack. Teachers are mandated reporters of suspected abuse or neglect in every US state, and both guards carry a hard stop for it: if your input suggests a reportable safeguarding concern, the pack stops drafting and points you to your school's reporting procedure and your state's law. A parent email, an incident summary, or a "let me just document this first" write-up is never the right response to a reportable concern, and drafting one first can cost time that matters. Report it the way your district trained you to, then come back to the paperwork.

The pack is built to use real, attributable sources — the revised Bloom's taxonomy to Anderson & Krathwohl (2001), Depth of Knowledge to Webb (1997), backward design to Wiggins & McTighe, UDL to CAST, formative assessment to Black & Wiliam (1998) — and to write "general convention" where a practice genuinely has no canonical author (exit tickets and bell-ringers, for instance) rather than inventing an authority for it. On standards specifically: every skill drafts to the framework you provide and marks alignment for you to verify. It will write [your standard here] rather than guess at a code, and it will never tell you a lesson is "standards-compliant" — that's a judgment your district makes, not a claim software gets to make. If you ever see a confident-looking code, statistic, or citation you can't source, don't ship it — ask the skill where it came from.

No, and you shouldn't want it to. AI writing detectors have documented false-positive problems, and accusing a student on a detector score is how teachers end up in an unwinnable meeting. Use /ai-resistant-assessment instead — it redesigns the assignment so a generic AI answer fails: process evidence, in-class components, personal application, oral defense. If you have a genuine integrity concern about a specific piece of work, it goes through your district's academic integrity procedure, on evidence, not through a score.

Yes. The pack ships a raw prompts pack: every prompt in click-to-copy HTML and PDF, plus one plain-text file per prompt. Paste them into ChatGPT, Claude, Gemini, or whatever your district allows — no install, no Project, no plugin. You give up the saved profile and the always-on guards, so you'll re-state your grade, subject, and standards each time, and the de-identification is entirely on you: strip student names before you paste. Worth knowing: claude.ai on the web is not supported for the plugin path, because the website doesn't run plugins. The Claude desktop app and Claude Code do.

You need to be in a conversation inside your Project, not a loose chat. Create one from Projects in the left sidebar, open it, start a new conversation there, and say set up my Job Search OS again (or type /jobsearch-setup-wizard). The Project is where your profile and trackers are saved.

That's the one field the wizard insists on — the salary guard can't run without it. If you genuinely don't want one, say "no minimum" and it records that the guard is off; roles below what you'd take will not be flagged until you set one. Paid hourly or on commission? Give the minimum in that unit ("$44/hour", "$55,000 base or $120,000 expected total") and the check compares like with like.

Work inside the Project where you ran setup. If you skipped setup, say set up my Job Search OS — paste your resume and it fills in most of your profile in two minutes. If setup ran but drafts still feel thin, the fix is almost always more wins and better voice samples: say "complete my profile".

Say connect my tools. It checks each connection and tells you what's missing (connectors are added under Customize → Connectors — use a personal account). Until then, jobs hand you paste-ready output and save files in your Project.

By design. A job search never runs through a current employer's mail or Drive. Connect a personal account, or paste the message.

It searches for recruiters, hiring teams, and anyone on your Application Board's Contact column, then shows you the list to confirm. Paste the missing message, or add the sender to the board's Contact column and run it again.

That's by design — nothing touches your email, calendar, or files without your click. Choosing Allow for this task covers the rest of that job. Tracker updates always ask with a quick confirm.

Open it from Artifacts in the sidebar to confirm it exists, then tell the job: "use the Application Board" (or whichever it is). Trackers named "SAMPLE — …" came from the sample run and are ignored on purpose.

Scheduled runs prepare everything and save it for your review — they don't read your inbox or write to your email or calendar on their own. Open the conversation it created, review, and say "move these to Gmail" or "add my focus blocks."

Ask for the larger data archive (Me → Settings & Privacy → Data privacy → Get a copy of your data) — that's the one with connections. LinkedIn emails a download link, usually within a day, and the link works for 72 hours. Unzip it and upload the connections spreadsheet (the file named Connections) to your Project. Until then, say "add a contact" to put people on your People board by hand.

On purpose. It only keeps people at companies on your Application Board — never your whole export — and it matches by company name, since LinkedIn usually leaves email addresses blank. If someone's missing, their company name may be written differently: say "who do I know at [company]" and confirm the possible matches.

Then it won't guess. Paste the posting and the company's about or careers page (or a recent article), and it builds the brief from those, with each fact's source.

That's honest, not broken: under about 10 applications from a source, a reply rate is noise. It fills in as the board grows. Blank Source or date columns on older rows count as unknown — say "fill in the sources" to add them.

The check lists signals, not verdicts — scammers copy real company names, and real employers sometimes do clumsy things. Find the role on the company's own careers site and check the recruiter's email domain; if it checks out, say "it checked out" and carry on.

Say "open my command center" from anywhere — it drops what was in progress and shows home again.

That's the point: a term the posting uses but your record doesn't show comes back as a question instead of going on your resume. Answer with what you actually did and it adds the line; say "no" and it stays off.

It shouldn't — every figure, title, and date is meant to trace to your resume or profile, and anything without a source becomes [confirm]. If one slipped through, say "where did this come from?" and give the real figure or tell it to cut the line. Never send a draft with an unconfirmed claim.

Say "applying anyway." It marks the row "below minimum — applying" on your board and carries on. The flag exists so a role below your minimum doesn't quietly become your main option.

Correct — it organizes evidence you supply or that public search returns, each figure with a source and date, and it never invents a "typical" rate. Run /comp-research and give it what you have.

Those get a "worth a professional's eyes" line on purpose. The scripts help you ask the right questions; an employment attorney or your jurisdiction's labor office answers them.

Output will be generic — it can't see your profile or trackers. Move to a conversation inside your Project and run it again.