Docs / Use cases

LP fundraising

Rainmaker is a fundraise chief of staff for a GP without an IR team. You meet a dozen qualified LPs at a conference and tell your agent about each one in a voice note. It stores them, drafts the follow-ups into your own Gmail in your own voice, arms the reminders, and every morning tells you who to chase next and why — including the ones your CRM says are fine but your notes say are dead. When you want to see the raise, the Board shows the whole book one card per firm, and behind every card an account brief for the whole relationship plus a pre-call brief per person, where every line cites the meeting, email or note it came from. Nothing is ever sent without you.

The loop at a glance

Dump · a voice note after each LPStore · deduped, staged, with the one unique thingDraft · into your Gmail, your voice, never sentReport · drafted, sent, replied, booked
↓
Digest · every morning: who, and whyTriage · the whole book, once a daySweep · replies and bookings, twice a daySync · your CRM, every night

A hosted cloud tool holds the memory and does the judging (flat credits per call); a free local skill does everything that touches your Gmail, calendar and CRM on your own machine. Every action the cloud suggests comes back with an outcome, or it resurfaces tomorrow.

Hand this page to your agent

AG2 Space 0.6.10 and newer. Send this page to your agent; no clicking through the Marketplace one tool at a time.

Copy this page's link from your browser and send it to your Sutando. It reads the page, installs every skill and activates every cloud tool listed under Requires below in one go, then follows the rest of the page with you, step by step. Before installing it shows you the plan: what it will add and what you already have. Anything that costs credits waits for your OK. Skills and cloud tools work the moment they land: your agent uses a newly activated cloud tool straight away, with no restart and no dashboard step. Only if the setup script itself reports that a restart is required does your agent ask you, once, to open Agent settings (the bot icon, bottom left), scroll to Runtime and click Restart engine.

Requires

  • Cloud tools: rainmaker
  • Skills: lp-pipeline, rainmaker-board

Or set it up by hand

Two pieces, two verbs, and they only work together:

  • Rainmaker — Activate. The hosted cloud tool. It keeps your LP memory (contacts, stages, check sizes, allocation timing, every interaction), runs your named follow-up workflows, and produces the daily digest. It never touches your inbox or your CRM.

  • LP Pipeline — Install. The free local skill. It runs inside your agent on your own machine: it transcribes and parses your voice notes, writes Gmail drafts, watches for replies and bookings, and connects to your CRM with your own credentials.

  • Rainmaker Board — Install. The free local skill that turns the memory into a page: the pipeline one card per firm, an account brief per firm and a pre-call brief per person, and buttons that hand work to your agent or write straight into your CRM. It also mints the private link that opens the same board on your phone.

Your agent restarts after each activate or install — that's normal. Once all three are in place, say "set up my LP pipeline" and it walks you through the rest in one sitting; the board is the last thing it opens, once the briefs exist. The skill ships a doctor: one command that checks every link of the chain (your Sutando id, Rainmaker reachable, fund facts, CRM connected and synced, the book, unreported actions, voice samples, the crons) and prints what is missing in the order to fix it. Setup is done when it says ready. Installed skills refresh on their own the next time you open the Marketplace; see how skills stay up to date.

Before you start

Four things your agent will ask for. Have them ready and setup takes ten minutes:

  • Fund facts. Fund name and target, your Calendly, the data-room link, and the deck — as the file in your Google Drive (your agent attaches the file itself), a local path, or a link only if that link is genuinely shared; a link nobody outside your firm can open is worse than no deck. For the Board it also wants the names of every fund and SPV you have raised (that is how it tells an existing investor from a prospect), your email domain (inbound versus outbound mail) and the team with their emails (owner names on cards). These ride along on every follow-up so you never paste them again.
  • Your Gmail, connected to your agent. Drafts are written into your Gmail; your agent needs the Gmail connection to do that (and to notice replies later). It creates drafts only — it has no path to send.
  • Three to five emails you actually sent to warm contacts. That is how it learns your voice. Without them the drafts sound like a bot, and an LP can tell.
  • Your follow-up styles, in your own words. If you already say things like "this one's a dodger" or "give him the padre treatment", those become named workflows (see Workflows you name). If you don't, the defaults are fine to start.

A CRM is optional. If you have one (day.ai, Affinity, HubSpot, Pipedrive, Attio), the opportunities in it become the starting book; if you don't, the voice notes are the book.

Step 1 · Set it up in one sitting

One conversation. Your agent does the setup; you answer questions and click one consent.

Start it

Set up my LP pipeline. We are raising [Fund name, target]. Deck: [link]. Data room: [link]. Calendly: [link]. My follow-up styles are: [describe two or three, in your words]. Pull five recent emails I sent to warm contacts as voice samples. Then tell me exactly which reminders and sweeps you armed.

It stores the facts, creates a workflow per style you described, saves the voice samples on your own machine, and arms the four crons on that machine: the morning briefing, two reply-and-booking sweeps a day, a nightly CRM sync that rebuilds the briefs, and a weekly portfolio-news scan. Then it runs the doctor and should end with ready. If it ends with anything else, the doctor's own list says what is still missing — a machine where the crons were never armed has no briefing and no sweeps, and the doctor says exactly that.

Connect day.ai (or another CRM)

Connect my day.ai to the LP pipeline. My pipeline there is called [Fund Raising]. Pull the opportunities into Rainmaker, tell me how many people came in, and which stage each landed in.

For day.ai your agent runs the official connection itself: your browser opens once, you sign in and click Authorize, and it does the rest. The token stays in a file on your own machine. For Affinity, HubSpot, Pipedrive and Attio it asks for your API key and stores it locally the same way — nothing CRM-related ever reaches AG2 Space. Expect most of a first sync to land in Identified: that is inventory nobody has met yet, and Rainmaker keeps it apart from real relationships on purpose.

Step 2 · After every LP: dump it

The one habit that makes this work. Thirty seconds, right after the conversation.

Send your agent a voice note in your AG2 Space room, or a memo on Slack, Telegram or Discord, or just type it. Name, firm, where you met, the one thing you'll otherwise forget. If you already know how you want to play it, say the codename.

The voice note (say it, don't type it)

Just met Joey Pimentel from the Pimentel family office in Monterrey, at SuperReturn. Does about ten million checks, only allocates in Q4, and hates decks longer than twelve pages. His daughter is starting a bakery. This one's a dodger.

That is a complete intake. From it your agent stores a contact with firm, geo, a $10M check band, a Q4 timing window and the two personal details; arms the "dodger" workflow; writes the day-0 follow-up into your Gmail drafts (the bakery goes in the first line, the deck goes in as a link, not an attachment, and it stays under twelve pages of anything); and reports back: created, draft waiting in Gmail with subject "…", day-3 nudge armed.

Ten LPs after a conference dinner is ten voice notes, or one long one. Duplicates are handled: the same person from two dumps, or from a dump and a CRM row, collapses into one contact (email first, then LinkedIn, then a fuzzy match on name and firm that copes with accents).

If a note produced nothing, your agent says so instead of staying quiet. A previous attempt at this — a folder of voice memos from Berlin that turned into zero drafts and no error — is the failure this whole product exists to prevent.

Step 3 · Your CRM is the source of truth

Opportunities in, never the 50,000-lead firehose. After a sync, Rainmaker shows exactly what day.ai shows.

LP Pipeline · CRM syncReads the opportunities in your CRM and the people on them — stage, expected commitment, close date, the notes on the deal — and imports them into Rainmaker with their source stamped. Closed-lost is dropped, the stage is mapped onto Rainmaker's own (Identified → identified, Initial Meeting → in dialogue, Soft Circle → soft-circled, and so on), the commitment becomes a check band and the close date a timing window. After a sync, every contact's stage equals the CRM's stage, down as well as up. A voice note can move someone forward between syncs; the next sync re-aligns them with the CRM, so a correction that matters belongs in the CRM, not in Rainmaker.
What a sync readsNot just the cards. For every organization with an opportunity it brings the activity the brief needs: every email thread with who sent it, who was on it, the subject and the message text your CRM exposes, so inbound and outbound are told apart; every meeting with its recorded summary and notes; the meetings on your calendar for the next sixty days; the open tasks; the CRM's own status line, next steps and warmth score for the organization; the opportunity's status, situation, challenges and probability; and each person's title and LinkedIn. The newest sixty threads and thirty meetings per organization are kept, all notes always.
Nameless rows are keptAn opportunity with an email and a firm but no person attached — a BCC recipient of your last raise announcement, a card nobody filled in — is not junk; it is a prospect. It comes in as Contact at <firm>, marked firm only: drafts to it open without a first name, and the placeholder is replaced the moment a real name shows up in a reply, a signature or the CRM. Names are never guessed from an email address. An organization with meetings and threads but nobody attached at all is kept too, as the organization's own row, so its history is never lost.
Cards filed under your own fundSome CRMs file a card created without an organization under your own firm. Those come in under the LP's real domain, never as your fund; your own team never appears as an LP; a mail provider is never a firm.
Coverage, verifiedAfter every import the skill compares what it pulled with what the cloud stored, organization by organization, and prints the gaps with the reason: an organization with no activity, a per-organization cap, a lookup the CRM refused. Nothing is assumed to have landed.

Sync it whenever the CRM changed

Sync my CRM into the LP pipeline and tell me what changed: how many new people, how many merged with someone I had already logged, and anyone whose stage moved.

The nightly cron does this on its own and rebuilds the briefs whose evidence changed. Ask for it by hand after a big day of updating the CRM, or before a conference.

The CRM is the source of truth. Rainmaker is the intelligence around your database, not a replacement for it: it never disagrees with the stage you see in day.ai after a sync, the daily triage tells you when your notes disagree with that stage, and the writes back into the CRM are the ones you make yourself — a stage move, a note, a reminder, a role tag, on the Board or by telling your agent — which it writes from your own machine with your own credentials, confirms by reading them back, and only then mirrors into Rainmaker. A refused write is shown to you, never hidden.

Rainmaker's own stages, in order: identified (inventory — a CRM row or a list entry nobody has met), met, followed up, in dialogue, meeting scheduled, soft-circled, committed; passed and existing LP sit to the side. Warmth is a separate thing from stage, and it is measured one way only: evidence of a recent two-way interaction. Fifty dormant rows under one firm name are fifty cold rows, not "strong institutional interest" — the triage is told so in as many words.

Step 4 · Every morning: "what should I do today"

Ranked, reasoned from your own notes, and honest about the state of the raise.

The briefing arrives on the morning cron, or whenever you ask. It has two halves, and the first one is the reason the product exists.

The triage. Once a day, before answering, Rainmaker re-reads the entire book — every contact, due or not — and comes back with a state-of-the-raise summary, the handful of priorities that matter most, suggested stage corrections, and hygiene flags. This is where "your CRM says Soft Circle but the note says they passed by email" surfaces. Stage corrections are never applied on their own: your agent puts each one to you as a one-line yes or no.

The actions. Then the ranked list: follow-ups due, next steps you promised in a note and haven't done, existing LPs due their quarterly touch, allocation windows opening in the next sixty days, portfolio news worth a personal note, and a few insights no rule could find — a whole batch from one conference gone quiet, a timing play, a relationship move. Every line carries its reason, written from your notes, not from a template: "send Fund III references — that was his one ask in São Paulo, and the $5M soft circle is waiting on it."

Ask for it any time

What should I do today?

Then work the list in the same conversation: "do 1, 3 and 4" and the drafts appear in Gmail; "dismiss 6, she told me she's out until January" and it never comes back; "move Maria to soft circle, she said yes on the phone" moves her in your CRM first, then syncs. Everything you accept, dismiss or complete is recorded — an action you ignore is back tomorrow, by design. That reporting is the whole accountability loop: if your agent drafts something and never reports it, tomorrow's briefing serves the same item again as overdue.

If the briefing says it ran on the fallback ranking rather than the model, the actions are still right — just less well reasoned. It says so out loud rather than guessing quietly.

Step 5 · Replies and bookings

Checked twice a day. A reply that waits until tomorrow's briefing is a missed one.

Your agent sweeps Gmail for replies from tracked LPs and watches for Calendly and calendar bookings that include one of their emails. A reply moves the person to in dialogue and cancels the remaining nudges of that workflow, so nobody gets a "just bumping this" after they already answered. A booking moves them to meeting scheduled — its own state, because a meeting on the calendar is more than a reply.

Tell it about a reply it could not see

Joey replied on WhatsApp, he wants a call next week. Record that and stop the nudges.

Anything that happened off-email — a WhatsApp, a hallway conversation, a call — you can just tell it. The stage moves, the workflow adjusts, and the next briefing reflects it.

When it's time to send

Show me the drafts waiting in Gmail for LPs and which ones are more than three days old.

Drafts are yours to send. This is the fastest way to see what is sitting there.

Step 6 · Portfolio news as a reason to write

"We just had a good Series C" is the most natural email you will ever send an LP.

Give your agent your portfolio companies once (they go into the fund facts). The weekly scan looks for material news — rounds, exits, milestones — and Rainmaker decides two things: whether the story is genuinely worth sharing, and which specific LPs get a personal note about it, by geography, entity type and what their notes say they care about. Never a blast; usually zero to four people per event.

Announce your own news

We just closed the Series C at [company] - share it. Tell me who should hear about it personally and draft those notes.

Your word is material; it skips the news search and goes straight to picking the right people and drafting.

Step 7 · The Board: the whole raise on one page

Your CRM shows deals. The Board shows relationships — and what to do about each one.

The Rainmaker Board: a white page with a stats row (open relationships, soft-circled and committed dollars, warm and cold counts, brief freshness), the state of the raise, a stage strip with whole-book totals, investor-type tabs, then the pipeline in the CRM's own stage columns with one card per firm showing amount, close date, days in stage, the people and the owner (sample data)
One board, one card per firm. Tabs split it by investor type; the columns are your CRM's own stages with the count and dollars per column; each card carries the amount summed across the firm, the close date, days in stage, the people at the firm, the owner, and three buttons: open the brief, draft the email, full history.

Say "show me the pipeline" and your agent renders it from Rainmaker's memory. What you get, top to bottom:

  • The state of the raise in one row — open relationships, dollars soft-circled and committed, who is warm right now, who has gone cold, and how many briefs are fresh. Under it, the triage's summary from this morning.
  • A stage strip with the whole-book totals — Identified 37 · Initial Meeting 24 · Due Diligence 10 … with dollars per stage. Each pill filters the board to that stage across every investor type.
  • Investor-type tabs — All types is one board; tap Family Office 27 · $47.1M and the board becomes just those.
  • The kanban itself, in your CRM's own stage names, with a colored bar and the firm count and dollars on every column. Columns open at ten cards and grow as you load more; search, owner, stage and temperature filters narrow it; the Table toggle lists every firm sortable by stage, amount, people, last touch and brief state.

Each card is one firm, however many people you know there: the firm, the lead person and their title, Existing investor or Prospect (any prior commitment to one of your entities, found automatically across every pipeline), the amount summed across the firm, the earliest close date, days in the current stage, the people at the firm, a temperature dot (from the last time they reached out: an email from them, a meeting, a reply; outbound mail never warms a relationship), and the owner. Five people at one firm were five cards before; they are one relationship now.

Open it

Show me the pipeline.

Your agent runs the preflight first (see Why the board comes last), renders the page, opens it, and tells you how many briefs are fresh, stale or missing. On the Mac the buttons act immediately; the same page from your phone queues the work for your agent (see On your phone).

Click a card: the account brief

The whole relationship on one page, then a tab per person.

An account brief sliding over the board: the firm name with badges, an Account tab first and one tab per person, a stat strip (amount across the firm, earliest close, days in stage, people, we owe, they owe), the account summary with evidence chips, stakeholders with role, influence and stance, open opportunities, and on the right we owe / they owe, risks, the merged interaction history and relationship data points (sample data)
The Account tab is the relationship: who the firm is to you across everyone you know there, who decides and who champions, the open opportunities, what each side owes, the risks, and one merged timeline. Every claim carries evidence chips; tap one to see the exact note, meeting or email behind it. The people tabs are the per-person briefs.

A firm where you know two or more people gets an account brief, written across all of them from the union of their evidence: a three-sentence account summary, the stakeholders with a role tag (economic buyer, champion, supporter, hurdle), how much they influence the decision and where they stand, the open opportunities with stage and amount, the decision drivers, we owe and they owe consolidated across people, the next steps ranked with the person to address and a drafted opener, the risks that could stall the account, and one merged interaction history with the person tagged on each line. A firm where you know one person opens on that person's brief. Until an account brief is built the page composes one from the people's briefs and says so.

The person tabs: the pre-call brief

Everything you need in the two minutes before the call — and every line tells you where it came from.

Each person tab is that person's own pre-call brief, the one to read before you write to them. It slides over the board so you never lose your place; Esc closes it, Back returns to the account, the arrows step through the firms in the column you were looking at.

The brief is written by Rainmaker from the person's record — the CRM fields, every note, every meeting summary, every email thread with its text, the upcoming meetings and open tasks your sync brought in — and it is held to one rule: every claim cites its source, or it is not shown. The model is handed a numbered list of evidence and must point at it; anything it says without a citation is dropped in code before you see it, and a "quote" that is not verbatim in the cited evidence loses its quote marks. What survives:

  • Header. Existing investor or Prospective investor first — the one thing anyone needs to know about a card in a glance — then investor type, stage, temperature and owner.
  • The stat strip. Prior commitment (the largest thing they ever did with one of your entities), the target for this fund, expected close and days out, days in the current stage, and the two mutual obligations: we owe them, they owe us.
  • Where we stand. Three sentences: who they are to you, what the last touch actually revealed in their words, and what is open in each direction.
  • What they flagged. The decision drivers, ranked by how much they gate the next stage, with their exact words when the record has them: "We come in bigger if we can ride along on two or three deals a year."
  • Deal notes surfaced from the CRM, suggested next steps ranked with a drafted opener for the top email, prior investments with your entities, the interaction history, everyone on this deal with a role tag (economic buyer, champion, hurdle, not active), and the relationship data points — where you met, their lens, the events they invited you to.
  • The footer says when the brief was generated and from how many meetings, emails and notes, and whether it is still fresh. A brief goes stale the moment its evidence changes — a new note, a reply, a stage move — and the page says so instead of pretending.

Prep for a call from the chat, too

Prep me for the call with [LP] this afternoon.

Five bullets, each citing the brief's evidence: where you stand, the drivers to address in their words, what you owe and are owed, the one question that moves the stage, the risk.

Buttons that do things

Every button hands the work to your own agent — and every CRM edit goes to your CRM first.

Draft the email · Draft follow-upYour agent writes the draft into your Gmail drafts, in your voice, from the top next step and its opener — never sent. The page shows the subject when it lands.
Prep me for a call · Schedule meetingA five-bullet prep in chat; or two or three slots from your calendar and a scheduling draft with your Calendly link.
Log a note · Set reminder · Move stage · Role tags · Investor typeThese write into your CRM — from your machine, with your credentials — and are read back to confirm before Rainmaker's copy is updated. If day.ai refuses (or your plan cannot do that kind of write), the page shows the exact reason. A stage move that no longer matches what the CRM shows is refused with a "refresh first" instead of overwriting someone else's change.
Refresh brief · Full historyRefresh pulls that one LP from the CRM again and rebuilds its brief and its firm's account brief (under a minute); Full history opens every interaction and note, newest first.
Every button, on your laptop or from the phone link, queues a task for your agent, which then acts from your machine. Nothing runs in the page and nothing sends. That is also why the link is the LP book: anyone holding it can read the board and queue draft tasks for your agent, so it belongs in your own DM.

On your phone, from your AG2 Space room

You are at a conference, not at your desk. The board comes to you as a link.

The same account brief on a phone: a full-screen sheet with a sticky bar (back, firm name, previous and next), the Account tab and the people tabs, the firm's badges, then the stat strip stacked (amount, earliest close, days in stage, people, we owe, they owe) and the account summary (sample data)
On a phone the brief is a full-screen sheet with a sticky back bar; the Account and people tabs scroll sideways; the stat strip stacks; the pipeline lists stages top to bottom; buttons are thumb-sized. Coming back from a brief lands you exactly where you were.

An AG2 Space room can carry a link, so that is how the board travels: your agent mints a private link to the board — one per Sutando account, revocable, never indexed — and posts it in your own room once. Open it on any phone and you get the live board rendered from Rainmaker's memory at that moment. Buttons on the hosted board queue work for your agent (the cloud never holds your Gmail or your CRM credentials): tap Draft the email on the train, and within a few minutes your agent has written the draft into your Gmail from your laptop and the page shows "Draft in Gmail: Fund IV deck and co-invest terms". CRM edits work the same way — queued, executed from your machine, confirmed.

Get the link

Share my board link here.

Your agent posts the link in the room you asked from. It refuses until the board is actually ready (briefs built), so nobody opens an empty page. If the link ever leaks, "rotate my board link" kills the old one.

Treat the link like the LP book itself: it belongs in your own room or DM, never in a shared channel. The page carries no names in its address, and nothing on it is indexed.

Why the board comes last

The board is the last step of a chain, and your agent checks the chain before it renders anything: Rainmaker reachable → your CRM connected → the opportunities synced (and how long ago) → the fund facts that the brief needs (your fund and entity names, your email domain, the team) → the briefs built. If a link is missing you get the exact next step, in order, instead of a half-empty board. The same check guards the phone link.

Every night the chain runs again on its own: the CRM's activity is pulled, the book re-synced, the coverage checked, and one build job rebuilds only the briefs whose evidence changed, people first, then the account briefs (a byte-identical sync rebuilds nothing). A 1,400-brief book takes about eighty minutes; your agent waits for the job and reports the counts. If a brief fails to build, the previous one stays and the card says brief failed; nothing stale is ever served silently.

Workflows you name yourself

A workflow is a follow-up playbook with days on it: a day-0 draft, a day-3 nudge, a day-10 breakup, a quarterly check-in. Rainmaker ships sensible defaults, but the ones that stick are the ones in your own vocabulary, because you will actually say them in a voice note at 11pm after a dinner.

Define one in your words

Create a workflow called 'dodger': day 0 send the follow-up with the deck link, day 3 a one-line nudge if no reply, day 10 a short breakup note. Skip the nudges if they replied. And 'existing LP': a check-in every quarter, never a pitch.

Steps that should stop after a reply are marked so; steps that should survive one (a quarterly cadence) are not. Retire a workflow any time and its pending steps stop materialising.

What it remembers, and where

Rainmaker's memory is keyed to your Sutando account, so it survives laptops, phones and forgetting: every contact with stage, check band, timing, cadence and source; the CRM's structured record of each opportunity (people and their roles, prior commitments, stage history); every interaction and outcome; every pre-call brief with its evidence; your fund facts; your workflows; the history of daily triages; and your board link. Ask for any of it in plain words — "where are we with the Monterrey family offices?", "who did I meet at LAVCA and never follow up with?", "what did the triage say last Monday?".

Your Gmail, your calendar, your CRM credentials and your voice samples stay on your own machine, inside the local skill. The cloud only ever receives what the skill has already parsed into a contact.

Guardrails

  • Nothing sends without you. Every follow-up is a Gmail draft. There is no send path.
  • Nothing dies silently. Every suggested action comes back with an outcome or resurfaces tomorrow; a voice note that produced nothing is reported as a failure, never dropped.
  • Your relationships, your channels. Your agent never contacts an LP on LinkedIn or a conference platform on your behalf. Existing LPs are touched on cadence, not on impulse.
  • Stage corrections are suggestions. The triage proposes; you confirm; the change is made in your CRM, then synced.
  • The CRM is the source of truth. After every sync Rainmaker's stages equal the CRM's, in both directions. The only writes into the CRM are the edits you make, done from your own machine with your own credentials, read back to confirm, and shown to you if refused.
  • Warmth is evidence, not arithmetic. A relationship is warm only on a recent two-way interaction; how many rows share a firm name means nothing.
  • Every claim in a brief cites its evidence. Uncited claims and non-verbatim quotes are removed in code before you see them; a brief whose evidence changed is marked stale, never served as fresh.

Credits & billing

Rainmaker's five cloud calls — logging contacts, importing a CRM batch, the daily digest, a portfolio-news scan, and building briefs (one call builds every stale person and account brief in the book as a job you poll for free) — each cost a flat, published credit price per call, whatever the size of the book; a call that fails its own checks is refunded automatically. Everything else is free: reading and correcting contacts, defining workflows, reporting outcomes, the source registry, the board page and your phone link. The LP Pipeline and Rainmaker Board skills are free and run on your own machine. Watch every debit and refund in your dashboard's wallet activity.

For your agent: the runbook

Hand this section to the agent that will run your raise. It is the whole job in order.

Everything below is what the LP Pipeline skill already tells the agent; it is here so you can see it, and paste it if an agent ever seems lost.

0 · Learn who you areCall the Rainmaker tool once (the digest, with triage skipped, is enough) and keep the Sutando user id it returns — every free endpoint needs it. The skill stores it for you.
1 · Run the doctor, do what it saysOne command lists what is missing and the next step for each, in order: fund facts (deck from Drive, Calendly, fund and entity names, domain, team), the CRM connection, the first sync, voice samples, the crons. Repeat until it answers ready. Never hand over a board before then.
2 · Connect and sync the CRMConnect day.ai (the user only clicks Authorize once). Run the probe once and keep its output. Pull the opportunities and their activity (nightly with the activity refresh, full when the doctor asks for it), fold them, upload for free, run the paid import batches the sync prints about a minute apart, then run the coverage check. Report created, merged and firm only counts and the coverage totals. After every sync the stages match the CRM.
3 · Arm the crons on this machineRegister the four entries the skill ships — morning briefing, reply and booking sweep twice a day, nightly CRM sync with the brief build, weekly portfolio scan — and tell the user which ones now exist. A machine where this never ran is silent, and the doctor will keep saying so.
4 · Every LP met: intakeParse the dump into contacts, store them in one call, execute every day-0 action it returns as a Gmail draft (the user's voice, the one unique thing first, deck attached from Drive, no dashes), then report each draft. Confirm fuzzy merges with the user. Never send.
5 · Every morning: digestOpen with the triage summary and hygiene lines, then the ranked actions with their reasons. Put each stage fix to the user as a yes or no and apply it in the CRM first. Report every action the user accepts, dismisses or completes — unreported actions return tomorrow.
6 · Twice a day: sweepReplies from tracked LPs and bookings with their email become interactions; a reply advances the stage and cancels the remaining nudges, a booking moves them to meeting scheduled. Say it out loud when someone books.
7 · The BoardDoctor first. Then ONE build-briefs call for the whole book (people and account briefs, as a job), wait for it, render, and tell the user how many person and account briefs are fresh, stale or missing. Never loop the build and never run two at once. Share the private link only in the user's own DM. Board edits go to the CRM first, are read back, then mirrored.
Rules that never bendDrafts only, never a send. The user's voice, never dashes. The CRM is the source of truth. Every claim cites. Every action is reported. A contact marked firm only is greeted without a first name. One build call per sync. Space the billed calls. Nothing dies silently — say what failed. Post one report at the end of a sync, not a running commentary.

If an agent is lost, say this

Read the LP Pipeline and Rainmaker Board skills, run the LP pipeline doctor, and do its next steps in order until it says ready. Then show me the pipeline.

That single instruction is the whole setup. The doctor is deterministic: the same book and the same machine give the same answer, so two agents will not disagree about what is missing. The full operating manual, with every exit code, expected output and known failure, ships inside the LP Pipeline skill.