anthropics/claude-for-financial-advisors/prospect-intake

Prospect Intake

Turn the pile of statements a prospect hands over into something usable — a clean summary they can read, a clean handoff the analyst team can build a full proposal from, and a warm note that keeps the prospect engaged while that happens.

This skill is the intake step, not the proposal itself. The full investment proposal (proposed allocation, expected outcomes, fees, transition plan) is built by the firm's analysts from the detailed handoff this skill produces — it is not generated here.

Inputs

Required: the prospect's name. Their files are needed too — either found in a connected tool or uploaded — but a request that arrives without them is a normal starting point, not a reason to hold off: Step 1: Find the Files below is how they get found, and it is where this skill begins. Never invent holdings.

Step 1: Find the Files

Ask the advisor a single question up front: do you want me to look for [prospect]'s files in your connected tools, or would you rather upload them here? Get the prospect's name either way — it's needed for both paths. Either way the prospect's files are confidential: read only the ones confirmed as theirs, and carry into the outputs only what those outputs need.

A. Look in connected tools

Check which systems are actually connected — the CRM (Redtail, Salesforce, Wealthbox), the firm's document store (for example @${CLAUDE_PLUGIN_ROOT}/partials/connectors/storage-examples.md), and email (for example @${CLAUDE_PLUGIN_ROOT}/partials/connectors/email-examples.md) — by looking for their tools the way the next paragraph says. Don't guess at a tool-name prefix, since real connector tool names use opaque prefixes.

Look for the tools before you trust the registry. Your own tools are the check that decides: a system whose tools are in your tool list, or come back from a ToolSearch by the system's own name, is connected and callable — use them. Where the session offers ToolSearch, a system's tools can be held back until it loads them, so search by name for every system above whose tools aren't already in your list; where it offers none, nothing is held back, and your tool list is every tool you have. ListConnectors can answer "No installed connectors found" even in a session with several live, working connectors, and in some clients it renders a user-facing card rather than returning data at all. So it explains a gap, it never establishes one, and an empty result means unknown, never "nothing is connected". When it does return entries, read enabledInChat, not connected: connected: true with enabledInChat: false is authenticated but switched off for this chat, so tell the advisor they can enable it here rather than reporting it as unconnected; a missing or null connected is unknown, not disconnected.

Search all connected systems in parallel — one system coming back empty doesn't block reading the others. The one exception is two CRMs. If tools come back for more than one CRM, both could hold this prospect's record and Step 5 writes to one of them, so follow the Ask-Once, Then Route Convention in its fixed order: search one CRM for the prospect, then ask the advisor once which CRM is the book of record, then search and write to only that one. Never search the second CRM before they answer, and never guess. A prospect not found in the first CRM is not a reason to search the second — a new prospect often isn't in either: ask then ("I didn't find [prospect] in Wealthbox — is Redtail where you keep prospects, or should this go into Wealthbox as a new contact?"). Before that first search, say which CRM you're looking in and that the book-of-record question comes as soon as it returns. The question comes when that first search returns, whatever it returned — a match, several candidates (they go in that same question), nothing at all, or an error you can't read — before any other call to either CRM. "It isn't in this one, so let me check the other" is exactly the search this rule forbids: which CRM holds the prospect is the advisor's to tell you, not yours to find out by searching. The document store and email aren't CRMs and are searched as usual. When they answer, offer to save the choice so they aren't asked again next intake — the offer opens the reply that takes their answer, before your first call to the CRM they named, even when their answer also hands you a different name to search: they paste "For new prospects, treat [CRM] as the book of record. Don't ask me to choose again." into Settings → General → Instructions for Claude — Claude can't write those settings, so never offer to save it yourself or say you have. If this pull is likely to take a moment (several systems connected), say so up front; that's a heads-up, not another "should I start?" gate — the upload-vs-search question above is the only required input before starting; the book-of-record question, when two CRMs are connected, comes during the search, not before it.

Show the advisor what matched (file names, email subjects, CRM record) before reading any of it — confirm which items are actually the prospect's files before pulling content from them. Disambiguate first: if a CRM search turns up more than one matching contact or opportunity, don't pull from or attach to any of them until the advisor confirms which one is correct — same privacy bar as a household match in other skills: never proceed on a name match alone.

If a meeting-intelligence tool (for example Zocks) is also connected, check it for a prior conversation with this prospect and pull whatever facts it already captured (goals, life details, account mentions). Reuse those instead of re-asking the advisor or the prospect for them, and cite them as coming from that conversation rather than treating them as missing in Step 2.

If no CRM, document store or email is connected, say so plainly and fall back to upload.

B. Upload here

Wait for the advisor to provide files — PDF/CSV/screenshot, or pasted text. Don't proceed on a promise of files to come; wait for them to actually arrive. Never invent holdings to fill a gap while waiting.

Step 2: Clean the Files

Dispatch one `claude-for-financial-advisors:statement-extract` subagent per file, all in a single message so the pile is read concurrently — one Agent(claude-for-financial-advisors:statement-extract) call per file, in the same response, each in the foreground (run_in_background: false) so every record arrives as its call's result inside this turn. Each gets one file path and the prospect's name, and returns that file's accounts as a normalized record with data-quality flags and its inferred-vs-read marking already done. Reading a stack of statements one after another is the slowest part of this skill, and it gets slower the bigger the pile — which is exactly when the advisor is waiting longest. A file whose dispatch errors, is refused because the concurrent-subagent limit is reached, or hasn't come back by the time you would otherwise end your turn is yours to read inline, to the schema below — never end your turn with a file outstanding.

Consolidation stays with you: the subagents each see one file and cannot tell that an account in file 3 is the same account as one in file 7. Duplicate detection across the set, and the single consolidated view, are yours to assemble from their records.

The extraction schema each subagent works to, which is also what to do by hand if the subagent is unavailable — per account:

• Account type/registration (401(k), IRA, brokerage, etc.) and custodian/provider if shown
• Holdings: ticker/fund name, shares or units, current value, cost basis if present
• Fees visible on the statement
• Statement/export date
• Any restrictions noted (e.g., employer plan in-service withdrawal rules, vesting)

Normalize everything into one consolidated view across all files/accounts. Flag data-quality problems as you go rather than silently working around them:

• Illegible or partial scans
• Missing cost basis
• Stale statement dates
• Accounts that appear duplicated across files

Never fabricate a figure that isn't legible or present — mark it "— not shown on file provided," unless it's a fact the meeting-intelligence tool already captured in a prior conversation (Step 1), in which case use that instead of flagging it missing, always labelled as from that conversation (Step 3).

Step 3: The Three Outputs

Write each as its own markdown file, clearly labeled — the prospect summary (A), the analyst handoff (B) and the what-to-expect memo (C) — using the templates in this skill's templates/ folder as the skeleton. Fill every field; where data wasn't shown on any file provided, mark it "— not shown on file provided" rather than leaving it blank silently.

Inferred vs. Read applies to every figure in all three outputs, not just the Analyst Handoff below — any number in the Prospect Summary or What-to-Expect memo that was computed, assumed, or taken from a prior conversation rather than read directly off a file needs the same flag.

A. Prospect Summary

Template: templates/prospect-summary-template.md. Plain-English, prospect-facing. Sections:

1. Accounts in Scope — account types/registrations reviewed (no account numbers, no custodian internals beyond what's needed to identify the account type)
2. Holdings Snapshot — a high-level view of what's held, by account
3. Total Assets Under Review — the sum across all accounts

No asset mix/allocation commentary, no fees, no performance or return claims — this confirms "here's what we received and understood," nothing more.

B. Analyst Handoff

Template: templates/analyst-handoff-template.md. Internal, for the firm's analyst team — this one stays internal (see Step 4). Sections:

1. Per-Account Line Items — the full consolidated detail from Step 2: account type/custodian, each holding with shares/units, value, cost basis, fees, statement date, restrictions
2. Data Quality Flags — illegible/partial scans, missing cost basis, stale statement dates, duplicate accounts across files, called out per account/line item
3. Inferred vs. Read — for every figure that required inference (e.g., computed from other numbers, assumed from context) rather than being read directly off a file, mark it as inferred; a figure from a prior conversation (Step 1) is marked as from that conversation; everything else is read-off-file by default

C. What-to-Expect Memo

Template: templates/what-to-expect-template.md. Brief, friendly, prospect-facing. Sections:

1. What We Received — the files/accounts reviewed
2. What Happens Next — the full proposal is now being prepared from this intake
3. Who You'll Hear From, and When — ask the advisor for the actual next point of contact and timeline; never invent a turnaround time, a name, or promise a specific outcome. If the advisor hasn't given you these yet, ask before finalizing this section rather than leaving a guess in place.

Step 4: Compliance

Run /compliance on outputs A (Prospect Summary) and C (What-to-Expect Memo) — both are prospect-facing and fall under the same marketing/antifraud rules as client communications.

Finish the review before you hand anything back. /compliance writes two files per document it reviews — <slug>-compliance-analysis.md and <slug>-compliance-redraft.md. Until both exist for both A and C, the review has not happened yet and the intake is not done. A clean review still writes a redraft, so "nothing needed changing" is not a reason for the file to be missing. Nothing here runs in the background: if you are waiting on a scan, it has already come back and the remaining markup is yours to write.