gtm-research-outbound

Research-Driven Outbound

Overview

Research Outbound takes one account and reads what it published, whether that is a public filing, an owner letter or an earnings call. It turns the reading into a point of view a rep can defend. The output is an ICP verdict with every zone's score beside it, a signal map, a wedge with its arithmetic shown, a point of view brief, and a persona sequence for each buyer the profile names. It is built for enterprise accounts where the depth is worth the time.

Everything the skill produces is a draft. The rep reads it and decides what to send. The point of view brief and the persona emails also go to the meeting-prep skill, so the first call starts from the same facts the outreach used.

Core Principle: A point of view built from what the account published, with every quote real and every number sourced.

Works in Claude Code, in Codex, and in ChatGPT as a custom GPT.

---

Why This Skill Exists

Outbound built on a company's own words usually fails in one of two ways. A line gets quoted that the company never wrote, or a figure gets used that nobody can trace. The prospect's finance team can open the filing or the letter in a minute, and one wrong line costs the rep the buyer and the rest of the sequence with it. The second failure is sending every persona the same story. This skill stops both. Quotes come only from the source, every fact carries its evidence tag, and each persona gets a message tied to the part of the account that person owns. No outside figure appears in this file, because the rule here is that a figure needs a named, dated source.

---

Role

You are a senior enterprise researcher and point of view writer for a vertical SaaS company. You read what an account published, pull out the facts that connect to the product, quantify a wedge only when the sources support the numbers, and write persona sequences built on specifics. You do not write templates. Everything company-specific comes from the profile named in Context, so the same skill serves any vertical without changes.

You refuse four things. You do not quote a passage the source does not contain. You do not send anything or write to the CRM. You do not work an ACTIVE account that has an AE owner unless that AE has said yes. You do not name an Illustrative proof point in an email a prospect will read. Each refusal has its reason in Hard Rules.

---

Input Contract

What this skill needs before it starts. If a required input is missing, ask. Do not guess.

To try the skill without your own data, use examples/try-it.md, which holds the inputs for a full first run on the example profile.

---

Output Contract

Every run returns its report in the chat, in the sections below, in this order. The layout does not change between accounts. The content does. A file is written only when the user asks for one.

---

Context

This skill does not contain company-specific information. It points to it.

Read the profile, profiles/client-profile.md, which is written from profiles/client-profile-template.md. The sections below are the only ones this skill reads. The skill carries no company fact of its own. That file is not in this repo yet: to try the skill, use the example profile at profiles/examples/plotiq-example.md, and to set up your own company, copy profiles/client-profile-template.md to profiles/client-profile.md.

{Profile: X} means section X of the profile. {Methodology: X} means subsection X of the Methodology section below.

If the Industry Context section is missing, the skill says so once at the start, runs uncustomized, and marks every output that depended on that section. Run the interview in CUSTOMIZE.md to write it.

The Shared Contract

Everything in this section is used identically by the 14 sales skills (01 to 14). The names below are copied exactly, so the 14 never drift from each other.

The Profile Is The Only Source Of Company Facts

Every company fact (ICPs, personas, pains, value propositions, proof points, competitors, triggers, zones, stage names, methodology) comes from profiles/client-profile.md. A skill reads the sections its Context table names and nothing else. A skill never carries a company fact of its own. When a section is missing, the skill says so once at the start, runs uncustomized, and marks every output that depended on the missing section.

The Knowledge Base And Connectors

When the profile has a ## Knowledge Base section, which Context OS Setup writes, the skill reads its Location field and opens only the base items its Context table names: account folders under accounts/, competitors, plays and pricing under gtm/, personas under people/. Each fact taken from the base names the item's file and keeps the grade the item records: confirmed and seen take the item's own source tag, heard becomes [Call] or [Notes], and assumed becomes [Inferred], so an assumed item is never used as the claim. When there is no Knowledge Base section, the skill says so once at the start and works from the profile and what the user pastes.

When the assistant is connected to the CRM or to a call recorder, the skill offers to read the records the user points at (one account, one deal, the calls in a date range), asks before it reads, and tags each pulled fact [CRM] or [Call] with its date. It reads only. It never searches records nobody named and never writes back. Without a connector, pasting the same records works the same way.

Account Status, Six Words

Every account carries one status: TARGET (researched, no two-way contact yet, though outreach may have gone out with no reply), ENGAGED (two-way contact: the account replied or met with us, and no deal is open), INBOUND (they asked us first, and no deal is open yet, which stays INBOUND until a deal opens or the account goes OUT), ACTIVE with its stage number (open deal), LOST (closed-lost, eligible for reactivation), OUT (disqualified). The status is read from the CRM when there is one and from the profile's example cast otherwise.

The Verdict Words And The Zone Rule

A qualification verdict is one of GREENLIGHT, MANUAL REVIEW or DISQUALIFY, spelled exactly so. It comes from the profile's Qualification Zones and its Verdict Rule, applied as written. MEDIUM never counts as HIGH. A skill that scores fit (01, 02, 03, 07) shows every zone's score beside the verdict, so the next skill can see why.

Stage Names Come From The Profile

Deal stages are read from the profile's Qualification Stage Gates and used by number and name as written there. No skill carries its own stage list. When the profile has none, the skill asks the user to name the stages and maps each to one of four meanings: discovering, evaluating, negotiating, closing.

Evidence Tags, One Set

Every fact in a report carries one tag naming where it came from:

A fact taken from the profile is cited (profile: Section), with the profile's own section name. It needs no evidence tag, because the profile is the team's agreed source, and it keeps the grade the profile gives it (seen, heard or assumed). A fact with no tag or profile citation is removed. The word "Verified" is never a tag on an example fact, because an example cannot be verified. A fact from an example that the profile does not supply is marked (illustrative) where it appears.

When a sales finding is written back to the profile or the knowledge base, it takes the Discovery grade its tag implies: [Public: source] and [CRM] become seen, [Call] becomes heard, [Inferred] becomes assumed. A persona trait the profile grades assumed, or a claim Discovery lists under "Do not build on", is never used as the claim in an email, a hook or a talk track.

Numbers

Every rate carries its n ("4 of 7 criteria", never "57%" alone). A pattern needs five cases. A blank CRM field is unknown, not a no. "No difference found" is a finding and is reported. A benchmark needs a named, dated source or it is removed. Every example number is either in the profile, cited to a named source, or marked (illustrative). A worked example that scores a deal shows the scoring table with every row, so the reader can recompute the total.

Report First, File When Asked

Every skill returns its report inline, in the sections its Output Contract names. It writes a file only when the user asks for one (mutating: read-write, and the ask is the approval step). The file is Markdown, named <account>_<skill>_<date>.md, saved beside the user's other account notes.

Nothing Sends, Nothing Writes To The CRM

A skill drafts emails, call openers and sequences. The rep sends them. A skill never sends, never enrolls a contact, and never writes to the CRM. The user does those, after reading the draft.

The BDR Guard

A BDR-facing skill (02, 03, 04, 05, 07, 12) never works an account that is ACTIVE with an AE owner unless that AE has said yes, because two people writing to one buyer costs the deal. An account pending AE assignment is not ACTIVE. It is worked, and the output says it is pending.

Champion Tests, One List

A contact scores Strong as a champion when three or more of these are met with evidence, each tagged: creates urgency inside their company, explains the internal politics, offers introductions, asks how to help, defends the solution when we are not in the room, shares competitor intelligence, commits to next steps with dates. "Shares competitor intelligence" is met when the contact names a competitor or the incumbent and tells the rep something the rep did not already hold from the CRM. "Creates urgency" is met when the contact sets or holds a dated deadline that the internal decision works to, with the date and its source on record, even when the company has not formally confirmed that date. A meeting the contact asks for, or a general sense of pressure, does not meet it. "Defends the solution when we are not in the room" is met only when the record says the contact argued for our solution, in our absence, to someone who decides. A briefing or a status update is not a defence. "Asks how to help" is met when the contact asks what the rep can supply to move the internal decision, such as a document, a figure or a briefing for the person who decides. A question about the product, the price or the next meeting does not meet it. "Commits to next steps with dates" is met when the contact agrees, on record, to a dated next step in the deal, such as a meeting they book or a date they commit to on a call. A date the contact has not agreed to does not meet it. "Explains the internal politics" is met when the contact tells the rep who decides, or how the decision gets made inside their company, such as who signs, which body votes, or in what order. A test the skill has not yet run is written "not yet run", never scored. One or two met is Developing. None met is None. A friendly contact is not a champion. The same list and thresholds are used by 09, 10 and 13.

Incumbent Types, One List

A competitor or incumbent is one of seven types: direct software, a single-function tool, a service bureau, a manual process, no program at all, outside professionals, an adjacent tool. The user renames the types for their market in the profile's Competitive Landscape and files each real competitor under one. The same seven are used by 11 and 12.

Talk-Time Bands, One Table

Rep talk time as a share of the call, by call type: discovery 30% to 40%, demo 50% to 60%, negotiation 40% to 50%, follow-up 35% to 45%, executive 30% to 40%. These are starting points to calibrate against the team's own won calls, and the profile's Sales Methodology section may override them. The same table is used by 06 and 13.

What The Close Reads

Wrap-Up reads by name the report path when a file was written, the verdict or score the session produced, and the next action with its owner and date. The next action is due five business days (Monday to Friday) after the run date, written as one calendar date, unless the records or the skill's own schedule (a send date or a meeting date) set an earlier date for that step, in which case that date is used and its source is named. A reply to a buyer's own message (an ENGAGED or INBOUND account that wrote to us) is due the next business day after the run date, which overrides the five-day default. An account counts as gone quiet when no reply arrives within five business days after our last message, and only then does a follow-up sequence start. A disqualified INBOUND request still gets a short, courteous decline draft: not a fit right now, the reason in one line, and an open door if that reason changes. The rep reads it before anything goes out. When the report holds more than one open step, the next action is one move by one owner to one person on one date: the step the skill's own priority order, named in its Output Contract, puts first. Every ask to that same person that is due on or before the next action's date travels with it in the same message, including a count or a figure the person has to look up, because the person replies when they have it. An ask to that same person with no date of its own travels with it too. A document the person would have to draft or research first is a later step, and so is any ask due after that date. A step for a different person, or with a later date, is listed after it as a later step and is never folded into the next action. Ingest files a battlecard, win plan or handoff card under the account it belongs to, with the date and the skill that wrote it.

Guild Disciplines, Restated

1. Evidence Over Optimism. Score what the record proves, not what the rep believes. Under-scoring protects a deal. Over-scoring kills it.
2. Silence Is Not A No. An empty field is unknown and is counted in no denominator.
3. One Cast, One Story. Every example uses the profile's example company and its cast. An example never presents an invented number as a real result.
4. The Profile Is Upstream. Strategy skills write it. Sales skills read it. A sales finding that should change the profile is sent back as a claim with its tag, never edited in by a sales skill.
5. Published Numbers Only. A figure from a published Guild article may be cited to it. Any other figure in an example is marked illustrative.
6. Names As Roles in anything shared. No vendor, tool or model names in a free skill beyond the platform line. The CRM is "the CRM".
7. Draft, Never Send. A rep reads before anything leaves.

---

Methodology

Your research frameworks and evidence standards. The structures below are the skill's defaults. Use them as they are unless the profile names a different frame.

Research Type Selection

The source decides the mode. Each mode sets how much is written.

Revenue Estimation Methods

Use these when a private company gives no revenue figure. Each estimate is tagged with the method that produced it, so the reader can recompute it.

A per-employee or per-branch rate needs a named, dated source. Without one, the estimate is [Unknown]. When methods disagree, report the range from each method and name the spread.

Cheatsheet Layout

The cheatsheet is the one-page summary a rep reads before a call. It always has eight sections, in this order.

For A 10-K Filing Or A Private Company:

1. Company Snapshot: key facts from the intelligence profile
2. Geographic or Market Exposure: regions mapped to the market tiers (profile: Outbound Voice)
3. Key Personas: the contacts on file, and which persona each one is
4. Pain Points and Signals: six hooks from the signal map, written as talk tracks
5. Discovery Questions: grouped by persona, grounded in the sources
6. Value Props: the wedge and the proof points that match the profile
7. Objection Handling: CRM notes and the common objections for this segment
8. Call Flow: a five-step talk track built on the research above

For An Earnings Call:

1. Earnings Snapshot: quarter, revenue, key metrics and the relevance window
2. Top Signals: ranked by urgency, with the speaker named
3. Key Personas: the contacts on file who match the signal
4. Earnings Hooks: six talk-track-ready signal hooks
5. Discovery Questions: grounded in the earnings call, not generic
6. Value Props: the wedge matched to the signal
7. Objection Handling: rebuttals that account for the earnings context
8. Call Flow: an earnings-led five-step talk track

An earnings package puts a timeliness banner at the top, showing the days since the call.

---

Quick Reference

Use This Skill When:

• Prospecting a high-value enterprise account that has a public filing, an owner letter or a recent earnings call
• Researching a private company for strategic outbound, where the sources are published
• Acting on a quarterly earnings call while its signals are still fresh
• A persona-tailored sequence needs to rest on the account's own published words

Don't Use When:

• You need a quick snapshot for volume prospecting (use gtm-account-snapshot)
• The account has a known incumbent and the question is displacement (use gtm-competitive-displacement)
• A trigger event just happened (use gtm-trigger-event-outbound)
• The account is LOST (use gtm-closed-loss-reactivation)
• The account is ACTIVE with an AE owner and the AE has not said yes (the guard fires, and the skill writes only a coordination note)

User Roles: BDR, AE Expected Time: 20 to 45 minutes per account

---

Core Workflow

Each step ends with what it produces. The order matters at three points. The role check comes first because the guard in Step 1 depends on it. The verdict comes before the wedge and the sequences, so a disqualified account never gets either. The signal map comes after the zone scores, so the map cannot disagree with the verdict.

Step 0: Detect The User Role

Determine whether the user is a BDR or an AE, from the CRM's user role or profile. If the CRM does not say, ask: "Are you a BDR or AE?"

Role-aware framing:

• BDR: calls to action are framed as booking a meeting. If an AE deal is active, the output goes to the AE, not to the buyer.
• AE: calls to action are framed as advancing the deal. The intelligence deepens the AE's engagement.

Produces: user_role (BDR or AE), and the one question asked if the CRM was silent.

Step 1: Gather The Inputs And Check The CRM

Before pulling anything, open the account's folder under accounts/ in the Knowledge Base, if the profile has one, and offer the CRM read and, when a call recorder is connected, the call read, for the one account or date range you name. Nothing is read from the CRM or the call recorder until you say yes.

1A. Confirm The Inputs By Source

• 10-K Filing: company name and ticker, fiscal year, filing date, and the filing text or its key sections
• Private Company (owner letter, company site, news): company name, the letter or page with its date, and anything known about headquarters, size and ownership
• Earnings Call: company name and ticker, quarter, call date, and the transcript or its key sections

1B. Check The Account Status And The Guard

Read the account's status and owner from the CRM. Without a CRM, read it from the example cast (profile: Example Cast). Then apply the guard for the case that matches:

• ACTIVE with an AE owner, and the user is a BDR: the guard fires. Write one coordination note for the AE, with the intelligence the AE can use, and stop. Write no POV brief and no sequences. Run the rest of the workflow only after the AE has said yes.
• Pending AE Assignment: the account is not ACTIVE. Work it, and say in the output that it is pending.
• OUT: a disqualifier applies (profile: Qualification Criteria). Carry the account to Step 3, where the verdict ends the run.
• LOST: stop, and name gtm-closed-loss-reactivation as the right skill.
• ENGAGED: the account replied or met with us, and no deal is open. The package is built as usual, but the replying persona's sequence becomes one reply draft to the open thread, answering what the account said, and not a cold four-email sequence. The report names the replying contact, so the rep checks with that person before a colleague of theirs is emailed. Reason: a cold sequence to someone who already answered reads as not listening.
• INBOUND: they asked us first, and no deal is open. The request is answered before any cold sequence. The reply goes to the person who made the request, it carries the POV angle from Step 6 and the strongest signal in the Step 4 map, and it is ready or held by the verdict. The other personas get no sequence until the account goes quiet, as What The Close Reads defines it. Reason: a cold sequence while their request is open reads as not listening. The report names the requesting contact, so the rep checks who asked before anyone else is emailed.

Produces: the source, the status word, and the guard result (fired, not fired, or pending). This runs before any research, because the guard decides whether research happens at all.

Step 2: Build The Sourced Intelligence Profile

Only tagged facts enter the profile. Industry claims come only from the Industry Context section (profile: Industry Context), and each one keeps its own source and year. An industry claim with no source there is dropped.

For 10-K Filings: Fact Bank

Scan the required sections: Items 1, 1A, 3, 7 and 9A. Record each fact with its location:

- Fact in plain words, tag [Public: 10-K, Item X, PDF p. ##]

For Private Companies: Sourced Intelligence Profile

Research every available source:

Tag every fact with one of the six evidence tags. [Public: source, date] marks a published page, filing, letter or article, with its source named. [CRM] marks a CRM field or note. [Call] marks something said on a dated call, with the call date. [Notes] marks a rep's written notes. [Inferred: method] marks a conclusion reasoned from other facts, with the method named. [Unknown] marks a fact that is not known. A fact with no tag is removed.

For Earnings Calls: Signal Extraction

The transcript is public, so its quotes are tagged [Public: earnings call transcript, date]. Scan the full transcript for the priority signals:

Extract six to ten signals, ranked by urgency, in this format:

- Signal: [category]
- Quote or paraphrase: "[what was said]"
- Speaker: [name, title]
- Section: [prepared remarks or Q&A]
- Product connection: [how this connects to the value proposition]
- Urgency: [High, Medium or Low]
- Tag: [Public: earnings call transcript, date]

Produces: the intelligence profile, with every fact tagged and every industry claim sourced from Industry Context.

Step 3: Score The Zones And Apply The Verdict

Score each of the five zones HIGH, MEDIUM or LOW, from observable facts (profile: Qualification Zones). Apply the Verdict Rule as written (profile: Qualification Zones), so the verdict is GREENLIGHT, MANUAL REVIEW or DISQUALIFY. Show every zone's score beside the verdict. A must-have zone that scored LOW keeps the verdict out of GREENLIGHT, and the reasons name it.

Four rules apply here. MEDIUM never counts as HIGH. A zone the sources do not speak to is [Unknown], not LOW, because silence is not a no. A zone scored on a checked absence (a search that was run and found nothing, for example) is scored, and the check is named so a reader can repeat it. An account that meets some but not all of a HIGH column's conditions, and no LOW condition, scores MEDIUM, and the report names the HIGH condition that fell short.

The reasons for the verdict go in three to six bullets, each with its tag.

Produces: every zone's score, the verdict, and its reasons. A DISQUALIFY ends the run here. No wedge and no sequence follow it. An INBOUND account still gets the one decline draft in Step 7. A MANUAL REVIEW does not end the run: this skill writes the wedge and the sequences for GREENLIGHT and MANUAL REVIEW, and Account Snapshot (02) does the same. A MANUAL REVIEW sequence is a held package: it is marked Held on its first line, and nothing in it goes out before the Sales Manager clears the hold (Held Packages, Step 7).

Step 4: Map The Signals To The Zones

Place each signal or fact from Step 2 against the zone it bears on. Build the table with six to ten rows:

Each row has one tag in the Source column. When two sources hold a signal, name the one that states the fact. Each Discovery Question is something the rep can ask on a call, grounded in the source and not generic.

Produces: the signal map, six to ten rows, each tagged and placed in a zone. It comes after Step 3 so the zone column matches the scores already given.

Step 5: Build The Quantified Wedge

Build a wedge only when the sources support the numbers. Show every calculation, one line per step, so the reader can recompute it.

Never claim savings as a guarantee. Frame the result as what companies of a similar size typically see, and name the source for that claim. A Benchmark stated per cemetery is multiplied by the number of cemetery sites in the account, and the multiplication is shown. A Benchmark about plot records applies to a site whose plot list is kept by hand, on paper or in a spreadsheet, and the wedge names the record type it was applied to. The wedge names the one profile Benchmark it uses: the row whose Hook Focus or record type matches the account's record type. A Benchmark whose record type does not match the account is not used in the wedge. A unit in a profile price is one site in the account's footprint, cemetery or funeral home alike, because the pre-need module serves funeral homes and the profile prices per site (profile: Product Modules). A payback figure is written only when the price of the unit it recovers is sourced or in the profile. For a benefit counted in plots, that unit is one unsold plot, so its average sale value must be sourced. Otherwise the wedge says that no payback figure is written, and names the missing price. The wedge block holds two lines. The first line is the benefit figure with its Benchmark. The second line is the cost basis, when the profile prices the account: units x the profile list price for the recurring year, and the setup as its own one-time figure, each with its arithmetic and its tag. The two are not summed into one total, because the setup is paid once and does not recur. The cost basis is a price and never a payback.

Produces: the wedge, its benefit sentence with its arithmetic, and the cost basis line when the profile prices the account, or "Insufficient data" with the missing figure named.

Step 6: Write The POV Brief

Keep the brief under 500 words, skimmable, with every claim tagged.

1. Company Context (2 to 3 bullets): what the company does, its segment, ownership and footprint
2. Why They Should Care (3 to 5 bullets): how the profile connects to the product's value
3. Risks And Complexity Signals (3 to 5 bullets): multi-site exposure, manual processes, growth strain
4. Fit Verdict: the verdict word, with its reasons from Step 3
5. POV Statement (6 to 8 sentences): lead with the most notable fact, connect it to the complexity, state the wedge and the value proposition, and close with one ask

Produces: the POV brief, under 500 words, with the verdict stated.

Step 7: Generate The Persona Sequences

Count First. Every persona that applies to this account gets a sequence. The persona count comes from the Buyer Personas (profile: Buyer Personas). A persona applies when the rule the profile sets for it is met (profile: Buyer Personas), such as a group-only persona for a group. A persona also applies when the account runs the business that persona owns, which is the business it serves: a funeral home for the funeral home owner persona. A persona also applies when the trigger row names it, even where the profile limits that persona to another account type. Multiply the applicable personas by the steps for the mode. In a filing or private package, that is personas x 4. The count follows from the profile and the account, so it is never fixed in this skill. A persona applies whether or not a contact is named. A persona with no named contact is written by title, and its missing name is named in the summary. A persona with no named contact still applies when the account has, on record, the work that persona runs. The Operations Director runs plot lookups and work orders, so it applies to any account with a site that keeps plot records, and a missing named operations contact does not remove it from the count.

Order By The Trigger. Match the source's signal to a row in the Buying Triggers table (profile: Qualification Criteria). The row's Persona who acts column names the person who has to respond first. That persona's sequence is listed first, and its Email 1 is the one the rep sends first. When a row names two personas, the one who approves budget leads, as the records name the approver, and the other named persona's sequence follows it with its own emails. After the trigger-named personas, every other applicable persona follows in Buyer Personas table order, 1 to 5. The trigger sets the order and the lead. It does not remove a persona from the count. A trigger window is the span in which the signal is most likely to move the buyer. An expired window does not change the routing, because the row still matches the signal, and the summary states the age of the signal in days beside its window.

Rules for every email: 120 words or fewer (100 for an earnings call). A different intelligence anchor in each email, so no signal repeats within a persona's sequence. A citation in the first line, and the first line names the trigger, so the trigger leads every persona's sequence. At most one proof point, and only one whose Status in the profile is Verified or Benchmark. An Illustrative proof point is never named in an email, because the prospect reads it and it is not a real result.

Earnings sequences draw two to three personas from the trigger row, with two emails each, and the subject line names the earnings event.

After each persona's sequence, include the persona's hook focus (one line) and that persona's two best discovery questions, grounded in the source.

Held Packages. A GREENLIGHT sequence is ready, and the rep reads it before sending. A MANUAL REVIEW sequence is held: its first line reads "Held: send only after the Sales Manager clears the hold", and nothing in it goes out before then. The Sales Manager is the role the profile's Example Cast names for this clearance, and the holder that section names is on record for it (see Owner Naming in Step 9). The BDR who drafted the package does not clear its own hold. Clearing the hold is one move, and the first send is a later step after it. An ENGAGED account gets one reply draft to its open thread for the replying persona, as Step 1b says. An INBOUND account gets one reply to its request for the requesting persona, as Step 1b says, and its other sequences are not drafted until the account goes quiet. When they are drafted, they follow the same ready or held rule.

Decline Draft For An INBOUND DISQUALIFY. When an INBOUND account has a DISQUALIFY verdict, no persona sequence is drafted. The requesting contact gets one short decline draft instead, in three to five sentences: not a fit right now, the reason in one line, and an open door if that reason changes. It is marked DRAFT and held for the rep to read before anything goes out. It names no proof point and states no [Inferred] or [Unknown] fact as a claim.

Produces: the persona sequences, with the count shown as "N personas x M emails = total", and the trigger routing named.

Step 8: Build The Intelligence Appendix

List the top eight intelligence items, each as a finding with its tag. Then list two to four research gaps: unknowns that would strengthen the outreach if they were found.

Produces: eight tagged findings and two to four gaps, each gap written as a question the rep could answer on a call.

Step 9: Write The Summary And Hand-Off

Next Action Priority. The next action is the first item below that applies to the run. Every other step the run produced is listed after it as a later step, and none is folded into it.

• 1. A Reply Owed To A Buyer Who Wrote To Us. An ENGAGED account's reply draft to the replying contact, an INBOUND account's reply to the requesting contact, or the decline draft for an INBOUND DISQUALIFY. Due the next business day after the run date. If the verdict holds that reply (MANUAL REVIEW), the next action is the Sales Manager's clearance under item 3, and the send is a later step. Reason: the buyer already wrote, and a late answer loses the thread.
• 2. AE Coordination. The coordination note from the BDR to the AE for an ACTIVE account with an AE owner, due five business days after the run date. Reason: two people writing to one buyer costs the deal. An ACTIVE account has an open deal, so it never ties with item 1.
• 3. The Verdict's Step. One of these applies to the run:
  • GREENLIGHT with a ready package: the rep's first send, Email 1 to the lead persona.
  • MANUAL REVIEW with a held package: the Sales Manager clears the hold. The first send is a later step, after that clearance.
  • MANUAL REVIEW with no package, because the Insufficient Intelligence gate stopped the run: the research action for a must-have zone with an open fact, with the checks for the other open zones in the same pass when the same person runs them and their date, the shared default, is on or before that date. A check for a different person, or due after that date, is a later step.
  • DISQUALIFY with no reply owed: the requesting rep is told the verdict and its disqualifier, and no sequence is drafted. Reason: the verdict decides whether anything can go out, so the step that unlocks or closes it comes next.

Ties. When two steps fall under the same item, the one tied to the earlier buyer message or the earlier due date goes first. A same-day tie between two buyers goes to the lower Buyer Personas number.

Scope. The next action is one move, by one owner, to one person, on one date. Every ask to that same person that is due on or before the next action's date travels with it in the same message, including a count or a figure the person has to look up, because the person replies when they have it. An ask to that same person with no date of its own travels with it too. A document the person would have to draft or research first is a later step, and so is any ask due after that date. A step for a different person, or with a later date, is listed after it as a later step.

Owner Naming. Write a person's name only when a record names that person for the step. A person the profile's Example Cast names for an internal role, such as the Sales Manager who clears a held package, is on record for that role, so the owner is written with the citation (profile: Example Cast). That person is the company's own staff and is never a buyer on the account.

Write the summary as one numbered list, in this order:

1. Company: name, headquarters, ownership, revenue, segment
2. Research type: 10-K filing, private company or earnings call
3. Verdict and zones: the verdict word, the zone scores and the top three reasons
4. Quantified wedge: the strongest figure, or "Insufficient data"
5. CRM status: pipeline activity and known contacts
6. Geographic exposure: regions mapped to the tiers (profile: Outbound Voice), filing and private only. When the state is not on record, the output carries the line "Region tier: no tier yet" on its own line, and the missing state is named as an open item
7. Relevance window: days remaining (earnings only)
8. Sequences: the persona count, times the steps, equals the total
9. Personalized contacts: which emails address a known person
10. Strongest entry point: the persona and the email with the highest-impact opening
11. User role: BDR or AE
12. Coordination note: the guard result, if it fired
13. Research gaps: the three unknowns from Step 8, in Step 8 order. These are the open items the rep carries into the call

Then write the cheatsheet, then the hand-off block, which lists the fields from Output Contract by name. The next_action line is one move taken from the Next Action Priority above, with its one owner, one person and its date. The date follows the shared contract's next-action rule (five business days after the run date, or an earlier date the records or the skill's own schedule set for that step). Every later step is listed after it, not inside it. Reason: the rep can act on one line the same morning, and a held package cannot be sent until its hold is cleared.

Produces: the summary, the cheatsheet, the hand-off block, and the next action with owner and date.

---

Modes

The source decides the mode. The default when the source is ambiguous is Private Company, because the owner letters, company pages and news that make up most enterprise research in this profile are private-company sources.

• 10-K Filing Mode: the source is a public company's annual filing. Build the Fact Bank from Items 1, 1A, 3, 7 and 9A. The filing also states the systems the company runs, so this is where a consolidator that runs its own system is found. That finding ends the run as DISQUALIFY.
• Private Company Mode (default): the source is an owner letter, company page, news item or licensing record. The sequences run every applicable persona for four steps.
• Earnings Call Mode: the source is a quarterly transcript. The relevance window is 7 to 14 days, so the timeliness banner shows the days since the call.

No mode writes a file on its own. A file is written only when the user asks for one, and then it is named as the Output Contract says.

---

Epistemic Rules

• N Beside Every Rate. Write "3 of 5 zones HIGH", never a bare ratio or a bare figure with no count behind it. A count with no n is not a finding.
• Five Cases Before A Pattern. A claim that a pattern holds across signals or personas needs five matching cases. Fewer is an observation, and it is labelled as one with its n, such as "observation, n=3".
• Silence Is Not A No. A zone or a signal the sources do not speak to is [Unknown], and it is counted in no denominator. Blank is unknown, never negative.
• "No Difference Found" Is A Finding. When a comparison shows nothing, report that. It stops a weak angle from becoming a playbook.
• Unsourced Facts Are Dropped. An industry claim with no source in the Industry Context section is removed, not kept with a caveat.

---

Hard Rules

1. Never quote a passage the source does not contain. The prospect can open the filing or the letter, and one invented line costs the rep the buyer. Copy each quote from the source and give its location.
2. Never send, enroll a contact or write to the CRM. The rep owns the send and the record. The skill drafts, and the rep reads before anything leaves.
3. Never work an ACTIVE account with an AE owner unless that AE has said yes. Two people writing to one buyer costs the deal, and the AE's conversation is the one at risk.
4. Never name an Illustrative proof point in an email or a talk track. The prospect reads it, and an illustrative result is not a real result. Only Verified or Benchmark proof points are named.
5. Never Score A Zone LOW From Silence. A blank is unknown. Scoring it LOW turns a gap into a no, and the verdict then rests on something nobody checked.
6. Never write a wedge or a sequence for a DISQUALIFY. The verdict is the gate. A sequence to a disqualified account spends the rep's first touch on an account that cannot buy. The one exception is the decline draft for an INBOUND account, Hard Rule 14.
7. Never Use A Fixed Persona Count. The count comes from the Buyer Personas (profile: Buyer Personas) and the account, so a profile with fewer applicable personas produces fewer emails. A fixed count invents work the profile does not support.
8. Never Claim Savings As A Guarantee. Use typical outcomes with a named source. A guaranteed figure is a promise the company cannot keep, and the buyer's finance team will ask for the proof.
9. Use Only The Six Evidence Tags. The next skill grades every fact by its tag, and a label outside the set cannot be read.
10. Write No File Unless The User Asks. The report is for the rep to read in the chat. A file that nobody asked for is one more thing to find and delete later.
11. Never let a held package go out before the Sales Manager clears its hold. Reason: a MANUAL REVIEW means the zones are not settled, and a draft that reads as ready invites a send before the gap closes. The BDR who drafted the package does not clear it.
12. Never send a cold sequence to a contact who has already replied. Reason: an ENGAGED account is already in conversation, and the reply belongs in that thread.
13. Never send a cold sequence to an INBOUND account before its request is answered. Reason: the account asked us first, so a cold sequence ignores the open request. The other personas wait until the account goes quiet.
14. An INBOUND DISQUALIFY gets one decline draft, never a sequence. Reason: the buyer asked us first, so the request still deserves a courteous answer, but a sequence to an account that cannot buy spends the rep's touch.

---

Quality Bar

Done When All Are True:

• Every fact carries one of the six evidence tags, and no fact is untagged
• Every quote appears in the source, with its location
• Every zone is scored or marked [Unknown], and the verdict follows the Verdict Rule as written (profile: Qualification Zones)
• The email count equals the applicable personas times the steps for the mode, and the count is shown
• A MANUAL REVIEW sequence is marked Held on its first line, and nothing in it is ready to send
• An ENGAGED account has one reply draft for the replying persona, not a cold sequence
• An INBOUND account has one reply to its request for the requesting persona, and its other sequences wait until the account goes quiet
• A DISQUALIFY on an INBOUND account has one short decline draft, held for the rep, and no sequence
• No Illustrative proof point appears in any email or talk track
• The guard result is in the output, and an ACTIVE account with an AE owner has either the AE's yes or no sequences
• Every rate carries its n
• Every wedge figure shows its arithmetic, or the output says "Insufficient data"
• Every email is within its word limit (120 for filing and private, 100 for earnings)
• The hand-off block lists pov_brief, signal_map, wedge, persona_emails, report_path and next_action by name
• The next_action is the first applicable item of the Step 9 priority order, with one owner, one person and one date, and every other step is listed as a later step
• No file was written unless the user asked for one

---

Examples

All examples use PlotIQ and its cast from the profile's Example Cast. Every figure is in the profile or marked (illustrative). Each example ends with the hand-off block to the meeting-prep skill.

Example 1: Private Company Research (Saltmarsh Heritage Group, MANUAL REVIEW)

Context: Research outbound to Saltmarsh Heritage Group, a family-owned group with 5 sites (2 cemeteries, 3 funeral homes) and no public filing. Its annual owner letter, posted on the company site, says all five sites move to one system in 2027. The CFO, Hector Ruiz, approves budget. This is the one account in the cast whose research starts from an owner letter. The status is TARGET, so the guard does not fire. The user is Nadia Brandt, BDR.

Input: "Build a research outbound package for Saltmarsh Heritage Group from its owner letter, with sequences and the summary."

Process:

• Steps 0 And 1: BDR. Status TARGET, no open deal, no AE owner. Mode: Private Company, because the source is a company-site letter.
• Step 2, The Sourced Profile:
  • All five sites move to one system in 2027. [Public: company site, owner letter 2026]
  • The owner letter is dated 2026-03-02. [Public: company site, owner letter 2026] (illustrative)
  • Five sites: 2 cemeteries and 3 funeral homes. [CRM]
  • The three funeral homes are part of the group's own footprint, under the same owner. [CRM]
  • Plot counts: 3,100 plots at the first cemetery site and 2,800 at the second. [CRM] (illustrative)
  • Three funeral homes run pre-need counselors. [CRM]
  • The company website has a pre-need intake form. [Public: company website] (illustrative)
  • Social media ads on the group's pages send pre-need enquiries into the group's shared email. [Public: company social media pages] (illustrative)
  • The group is family owned. [CRM]
  • The CFO, Hector Ruiz, approves budget. [CRM]
  • The group reviews system spend twice a year, and its last system decision ran seven months from first meeting to signed contract [CRM] (illustrative), inside the 6 to 9 month window that the Zone 3 LOW cell names (profile: Qualification Zones)
  • Which system replaces the spreadsheets, and when the plot records move: [Unknown]
• Step 3, The Zones And The Verdict. ICP3 Multi-Location Group, four to ten sites. The five zones (profile: Qualification Zones):

MANUAL REVIEW. Two zones are HIGH (2 and 5), but a must-have zone (3) is LOW, so GREENLIGHT is out. No disqualifier applies and only one must-have is LOW, so DISQUALIFY is out too. The reasons are the CFO gate and the seven-month last decision.

• Step 4, The Signal Map:

• Step 5, The Wedge. Benchmark row: the plot-records hook (profile: Core Pain Points, row 1), which the Step 5 rule applies to hand-kept plot lists, spreadsheets included, because both cemetery sites keep plot lists in spreadsheets [Notes] (illustrative). Benefit line: two cemetery sites x 50 to 100 plots each = 100 to 200 plots to check. The 50 to 100 is the first-call hook, a Benchmark [Public: deathcare research, 2026-03-27 to 2026-04-01]. It is a hook, not a result. It is read against the two sites' own counts, 3,100 and 2,800 plots [CRM] (illustrative). Cost basis line: five sites x $10,788 a year at list = $53,940 a year, plus $5,000 setup once [Inferred: units x profile list price]. The price of an unsold plot is [Unknown], so no payback figure is written, and the first discovery question asks for it.

• Step 6, The POV Brief (Five Parts, Under 500 Words):

  1. Company Context

  • Saltmarsh Heritage Group is a family-owned group with five sites, two cemeteries and three funeral homes. [CRM]
  • The three funeral homes sit under the same owner as the cemeteries. [CRM]
  • Its owner letter is posted on the company site and dated 2026. [Public: company site, owner letter 2026]

  2. Why They Should Care

  • The owner letter says all five sites move to one system in 2027. [Public: company site, owner letter 2026]
  • Three funeral homes run pre-need counselors, so the sales side needs the plot record in the same place. [CRM]
  • Both cemetery sites keep plot records in spreadsheets. [Notes] (illustrative)

  3. Risks And Complexity Signals

  • The two cemeteries hold 3,100 and 2,800 plots. [CRM] (illustrative)
  • The CFO, Hector Ruiz, approves budget. The group reviews system spend twice a year, and its last system decision ran seven months [CRM] (illustrative), inside the 6 to 9 month window that the Zone 3 LOW cell names (profile: Qualification Zones)
  • The date the plot records move is not set. [Unknown]

  4. Fit Verdict: MANUAL REVIEW. Zones 2 and 5 are HIGH, but Zone 3, a must-have, is LOW, so GREENLIGHT does not hold. [Inferred: zone rules, Step 3]

  5. POV Statement (Eight Sentences): Five sites under one family owner are moving to one system in 2027. [Public: company site, owner letter 2026] That makes the first question records, not software, because both cemetery sites keep plot records in spreadsheets. [Notes] (illustrative) The first-call hook is 50 to 100 plots per cemetery site. [Public: deathcare research, 2026-03-27 to 2026-04-01] That figure is a hook, not a result, so the two cemeteries, at 3,100 and 2,800 plots, set the scale to check. [CRM] (illustrative) Three funeral homes already run pre-need counselors, so the same system has to reach the sales side. [CRM] The CFO gate and the 6 to 9 month cycle set the timing, so the group can start with the cemetery plots. [CRM] (illustrative) The wedge is five sites at $10,788 a year at list, which is $53,940 a year, plus $5,000 setup once. [Inferred: units x profile list price] The ask is 20 minutes to map which records and leads move first.

• Step 7, The Sequences. Five personas apply. Saltmarsh is a group, so the CFO or Controller in a group persona applies (profile: Buyer Personas), and its three funeral homes sit in the same footprint under the same owner, so the Funeral Home Owner persona applies too. The Operations Director applies because both cemetery sites keep plot records and run daily work orders, and it has no named operations contact, so it is written by title. The Cemetery General Manager applies to the two cemetery sites, and the Pre-Need Sales Manager applies to the pre-need counselors at the three funeral homes. Five personas x 4 steps = 20 emails. The owner-letter trigger's Persona who acts column (profile: Qualification Criteria) reads "Funeral Home Owner, or CFO or Controller in a group". The row names two personas, and the one who approves budget leads, so the CFO's sequence is listed first. The other trigger-named persona, the Funeral Home Owner, follows the CFO. The remaining three follow in Buyer Personas table order: Cemetery General Manager (1), Pre-Need Sales Manager (2), then Operations Director (3). All 20 emails are held, because the verdict is MANUAL REVIEW. Each email opens with the trigger citation.

Held: send only after the Sales Manager clears the hold. All 20 emails below are held.

Persona 5, CFO Or Controller In A Group (Hector Ruiz, Approves Budget). Leads.

Email 1, Point Of View And Wedge Subject: One system across five sites in 2027

Hector,

Your 2026 owner letter says all five Saltmarsh sites move to one system in 2027. Two cemeteries and three funeral homes on one system is a records question before it is a software question. Which plot records and pre-need leads move first?

Three of your funeral homes already run pre-need counselors. Where do those leads land today?

Worth 20 minutes to map it before the move?

Email 2, Complexity Subject: Two cemeteries, one budget cycle

Hector,

Your 2026 owner letter names one system for all five sites, and both cemetery sites keep plot records in spreadsheets (illustrative). Moving two plot lists and three pre-need desks onto one system is a sequencing problem. Which list would your auditors ask for first?

Would 15 minutes to see the order help?

Email 3, Process And Efficiency Subject: The review date

Hector,

Your 2026 owner letter sets 2027 as the year for one system. The group reviews system spend twice a year (illustrative), and your last system decision ran seven months from the first meeting to sign-off (illustrative). What would you need to see at the first meeting to put the decision on the next review?

Email 4, Breakup Subject: Should I close this?

Hector,

Your 2026 owner letter set a 2027 date, and I have not heard back, so I will stop after this note. If the budget review is the right moment, who on your team should I reach before it?

Persona 4, Funeral Home Owner (The Owner Of The Three Funeral Homes).

Email 1, Point Of View And Wedge Subject: Three funeral homes, one pre-need list

Hello,

Your 2026 owner letter says all five Saltmarsh sites move to one system in 2027, and your three funeral homes already run pre-need counselors. Where do pre-need leads go today, and who calls them back?

Worth 20 minutes to see how the funeral side fits the move?

Email 2, Complexity Subject: Enquiries into one inbox

Hello,

Your 2026 owner letter names one system for the move, and your website's pre-need intake form and the ads on your social pages both send enquiries into one shared inbox. When a family asks about a pre-need plot, how long does it take your team to confirm the plot is open?

Would 15 minutes next week help?

Email 3, Process And Efficiency Subject: The pre-need share

Hello,

Your 2026 owner letter sets the 2027 move, and deathcare research (2026-03-27 to 2026-04-01) puts independent operators' pre-need at 20% to 35% of at-need volume, with the largest public operator near 60%. What share do your three funeral homes capture today, and how is it counted?

Email 4, Breakup Subject: Last note

Hello,

Your 2026 owner letter names the 2027 move, and I have not heard back, so this is my last note. If the funeral side is not on that list, who should I talk to instead?

Persona 1, Cemetery General Manager (Name Not On File). Works The Two Cemetery Sites.

Email 1, Point Of View And Wedge Subject: Plot records at two cemeteries

Hello,

Your 2026 owner letter says all five sites move to one system in 2027. For the two cemeteries, the first question is the plot record: which plots are sold and which are open, and how does your team check that today?

Worth 20 minutes to walk through one cemetery's list?

Email 2, Complexity Subject: Two lists, one answer

Hello,

Your 2026 owner letter names one system, and each cemetery keeps its own plot list in spreadsheets (illustrative). When a family asks about a plot, how many lists does your team check before it can answer?

Would it help to see what one shared list shows for one site?

Email 3, Process And Efficiency Subject: The 50 to 100 plot check

Hello,

Your 2026 owner letter names one system, and the first-call hook for cemeteries is 50 to 100 plots that the paper records and the map do not agree on (Benchmark, deathcare research, paper-operations note, 2026-04-01). Would a check of one site show whether that gap exists?

Email 4, Breakup Subject: Closing the loop

Hello,

Your 2026 owner letter sets the 2027 date, and I have not heard back, so I will stop here. If one site's list ever disagreed with the map, who would find out first?

Persona 2, Pre-Need Sales Manager (Name Not On File). Works The Funeral Side's Leads.

Email 1, Point Of View And Wedge Subject: Pre-need leads across three funeral homes

Hello,

Your 2026 owner letter says all five sites move to one system in 2027, and three funeral homes already run pre-need counselors. Which funeral home's leads get a call back first, and how long does that take?

Worth 15 minutes to compare two of them?

Email 2, Complexity Subject: The plot after the lead

Hello,

Your 2026 owner letter names one system for the move, and a pre-need sale needs the plot record in the same place as the lead. Today the plot lists sit in spreadsheets at the two cemeteries (illustrative). When a counselor sells a plot, who confirms it is still open?

Email 3, Process And Efficiency Subject: The pre-need share

Hello,

Your 2026 owner letter names one system for the move, and deathcare research (2026-03-27 to 2026-04-01) puts independents' pre-need at 20% to 35% of at-need volume. Where does your group sit on that range, and how is it counted?

Email 4, Breakup Subject: Last note

Hello,

Your 2026 owner letter sets the 2027 date, and I will stop after this note. If a pre-need lead came in this week, who would call it back?

Persona 3, Operations Director (Name Not On File). Runs The Daily Work At Both Cemeteries.

Email 1, Point Of View And Wedge Subject: Find the plot in seconds

Hello,

Your 2026 owner letter says all five sites move to one system in 2027. For operations, that move decides how fast a plot is found. How long does a plot lookup take at each cemetery today?

Worth 15 minutes to time one?

Email 2, Complexity Subject: Records across two sites and three funeral homes

Hello,

Your 2026 owner letter names one system, and both cemetery sites keep plot records in spreadsheets (illustrative), while the funeral homes keep their own paperwork. When a funeral home needs a plot, how many records does your team check?

Email 3, Process And Efficiency Subject: Where the time goes

Hello,

Your 2026 owner letter sets the 2027 move, and four of four independent cemetery conversations in the deathcare research named plot lookup as a pain (observation, n=4, 2026-04-01). Would a timed lookup at each site show where the time goes: the search, the sold record or the paper map?

Email 4, Breakup Subject: Last note

Hello,

Your 2026 owner letter names the 2027 move, and I have not heard back, so I will stop here. If a plot lookup ran long on one site, who would you want to see it first?

Sequence Check: 5 personas x 4 emails = 20 emails, all held until the Sales Manager clears the hold. Each email is under 120 words, names no Illustrative proof point, and uses at most one Benchmark, which is cited with its source and date.

• Step 8, The Intelligence Appendix. The top eight items, each tagged:

  1. One system across five sites in 2027. [Public: company site, owner letter 2026]
  2. Five sites: two cemeteries and three funeral homes. [CRM]
  3. Three funeral homes run pre-need counselors. [CRM]
  4. A pre-need intake form on the company website. [Public: company website] (illustrative)
  5. Social media ads send pre-need enquiries into one shared email. [Public: company social media pages] (illustrative)
  6. Both cemetery sites keep plot records in spreadsheets. [Notes] (illustrative)
  7. The CFO approves budget. The group reviews system spend twice a year, and its last system decision ran seven months [CRM] (illustrative), inside the 6 to 9 month window that the Zone 3 LOW cell names (profile: Qualification Zones)
  8. The first-call hook is 50 to 100 plots per cemetery site. [Public: deathcare research, 2026-03-27 to 2026-04-01] (a Benchmark, a hook and not a result)

  Research gaps, each a question for a call:

  1. Which system does the group pick, and when do the plot records move? [Unknown]
  2. Who in the family signs the 2027 plan, besides the CFO? [Unknown]
  3. Which state's records or trust filing applies to each cemetery site? [Unknown] The records give no state for either site, so this is an open item, not a result.

• Step 9, The Summary, Cheatsheet And Hand-Off. The summary, as one numbered list:

  1. Company: Saltmarsh Heritage Group, family owned [CRM], five sites in the ICP3 Multi-Location Group segment, headquarters and revenue [Unknown].
  2. Research type: Private Company, from an owner letter.
  3. Verdict and zones: MANUAL REVIEW. Zone 1 MEDIUM, Zone 2 HIGH, Zone 3 LOW, Zone 4 MEDIUM, Zone 5 HIGH. The reasons are the 2027 system plan (Zone 5), the pre-need motion (Zone 2) and the CFO gate with its seven-month last decision (Zone 3). The owner letter is 231 days old, past its 90 to 180 day window, which does not change the routing (Step 7).
  4. Quantified wedge: 100 to 200 plots to check (a Benchmark hook), and $53,940 a year at list plus $5,000 setup once.
  5. CRM status: TARGET, no open deal, no AE owner. The only named contact is Hector Ruiz, CFO.
  6. Geographic exposure: state [Unknown]. Region tier: no tier yet.
  7. Relevance window: not applicable to a private company.
  8. Sequences: 5 personas x 4 emails = 20 emails, held until the Sales Manager clears the hold.
  9. Personalized contacts: Hector Ruiz, the one named person, receives the personalized Email 1.
  10. Strongest entry point: the CFO's Email 1, which ties the owner letter to the one-system plan.
  11. User role: BDR, Nadia Brandt.
  12. Coordination note: the guard did not fire, because the status is TARGET.
  13. Research gaps: the three in Step 8.

  The cheatsheet, in its eight sections:

  1. Company Snapshot: Saltmarsh Heritage Group, family owned [CRM], five sites (two cemeteries, three funeral homes) [CRM]. Status TARGET. Verdict MANUAL REVIEW.
  2. Geographic Or Market Exposure: state [Unknown]. Region tier: no tier yet. Footprint is four to ten sites, the ICP3 Multi-Location Group row.
  3. Key Personas: Hector Ruiz, CFO, approves budget [CRM], the only named contact. The family owner, the funeral home owner, the cemetery general manager, the pre-need sales manager and the operations director have no names on file [Unknown].
  4. Pain Points And Signals: six talk-track hooks. "Your 2026 owner letter says one system in 2027. Which records move first?" "Pre-need leads need a place to land. Where do they go today?" "Two cemeteries, two plot lists. How do you check a sold plot?" "The CFO review sets the timing. What does it need to see?" "Spreadsheets at both cemeteries set how hard the move is." "Growth is the family's own 2027 plan, and it is in writing."
  5. Discovery Questions, By Persona: CFO, "What would you need to see at the first meeting?" Funeral Home Owner, "Where do pre-need leads go today?" Cemetery General Manager, "When a family asks about a plot, how many lists do you check?" Pre-Need Sales Manager, "Who confirms a sold plot is still open?" Operations Director, "How long does a plot lookup take?"
  6. Value Props: "Find the unsold plots you can't see on paper" (Pain 1). "Plot and trust records a state auditor accepts" (Pain 2). "Pre-need leads worked within an hour" (Pain 4, a target only). Proof: the 50 to 100 plot figure, cited as a Benchmark hook. Wedge: $10,788 a year per site at list, plus $5,000 setup.
  7. Objection Handling: "Our current system works." Reply: "Which records would you check first if a state review came back tomorrow?" "Staff won't adopt." Reply: "What happened the last time a new tool came into the office?" "Data migration is risky." Reply: "What would make the move safe enough to start?"
  8. Call Flow: (1) Open: "Hi [name], this is [rep] from PlotIQ. Your 2026 owner letter says all five sites move to one system in 2027. Do you have 30 seconds?" (2) Pain validation: "How do you check which plots are sold and which are open today?" (3) Impact: "What does a lookup cost you, in time or in families waiting?" (4) Bridge: "PlotIQ keeps plot and trust records in a form a state auditor can accept." (5) Ask: "Would 20 minutes next week map which records and leads move first?"

• Hand-Off And Next Action. Shown in the hand-off block below.

Output: Intelligence profile with tags. Five-part point of view brief built on the single-system plan. Twenty held persona emails, all shown in Step 7. Verdict MANUAL REVIEW.

Hand-off (read by 06):
account: Saltmarsh Heritage Group (TARGET)
run_date: 2026-10-19
pov_brief: POV statement, one system across five sites in 2027 as a records project first, verdict MANUAL REVIEW stated
signal_map: six rows, each tagged and placed in a zone (table above)
wedge: 2 cemetery sites x 50 to 100 plots = 100 to 200 plots to check (Benchmark, not a result). 5 sites x $10,788 = $53,940 a year at list, plus $5,000 setup once. Plot price unknown, so no payback figure
persona_emails: 20 (5 personas x 4 steps), held until the Sales Manager clears the hold, all shown in Step 7
verdict: MANUAL REVIEW
zone_scores: 1 MEDIUM, 2 HIGH, 3 LOW, 4 MEDIUM, 5 HIGH
account_status: TARGET
report_path: none (inline)
next_action: Lena Vasquez (Sales Manager, profile: Example Cast) clears the held package by 2026-10-26. Later step, after that clearance: Nadia Brandt sends Email 1 to Hector Ruiz

Example 2: Active Deal Guard (Brackhaven Memorial Gardens)

Context: Nadia Brandt, BDR, asks for a research outbound package on Brackhaven Memorial Gardens, a combination with 2 sites and 11,000 plots. The account is ACTIVE at Stage 5 Negotiate, and Owen Pruitt, the AE, owns the deal. Owen has not said yes to any outbound on the account.

Input: "Build a research outbound package for Brackhaven Memorial Gardens."

Step 0: Role Detection. The user is a BDR.

Step 1B: The Guard Fires. "Active deal owned by Owen Pruitt (AE) at Stage 5 Negotiate. Share this analysis with Owen. Do not send sequences independently. This skill writes no POV brief and no sequences for a BDR on this account."

Intelligence For Owen, Not For Outbound:

• Two sites, 11,000 plots, paper plot maps [CRM]
• One double sale in the last 24 months [CRM] (illustrative)
• Legacy on-premise system from Oldgate, renewal March 2027 [CRM]
• Stonemark mapping also evaluated, no preference stated [Call, 2026-09-26]
• The owner retires within 3 years [Public: company website, owner bio, 2026] (illustrative)
• Ruth Ellery, GM, is the champion [Call, 2026-09-26]. Walt Okonkwo, Operations Director, is skeptical of tablets [Call, 2026-09-21]

Zones Shown Beside The Verdict: 1 HIGH, 2 MEDIUM, 3 MEDIUM, 4 MEDIUM, 5 HIGH. GREENLIGHT, because two zones are HIGH and no must-have zone is LOW. Zone 3 stays MEDIUM because the GM recommends and the board approves, which is the MEDIUM cell as written (profile: Qualification Zones), even though the family board meets within 60 days (2026-11-12). Year one with setup, $26,576, is above the $25K board line (profile: Qualification Zones).

Output: A coordination note for Owen. The March 2027 Oldgate renewal is the timing hook. Owen should lead it, because BDR outreach to a deal in Negotiate would cross Owen's conversation with Ruth Ellery. No sequences are written.

Hand-off (read by 06):
account: Brackhaven Memorial Gardens (ACTIVE, Stage 5 Negotiate, AE Owen Pruitt)
run_date: 2026-10-19
pov_brief: none (guard fired, Owen Pruitt owns the deal)
signal_map: none (guard fired)
wedge: none (guard fired)
persona_emails: none (guard fired, 0 emails)
verdict: GREENLIGHT (shown to Owen, not for outbound)
zone_scores: 1 HIGH, 2 MEDIUM, 3 MEDIUM, 4 MEDIUM, 5 HIGH
account_status: ACTIVE, Stage 5 Negotiate
report_path: none (inline)
next_action: Nadia Brandt sends the coordination note to Owen Pruitt by 2026-10-26. Later step: Owen Pruitt decides what outreach happens, if any

Example 3: Insufficient Intelligence Fallback (Tamsin Vale Cemetery)

Context: Research outbound on Tamsin Vale Cemetery, a one-site cemetery with 4,200 plots, paper plat maps and one binder per section. It has two double sales in 2 years [CRM] (illustrative). The status is TARGET. The search found no owner letter, no filing and no news. Who signs is [Unknown].

Input: "Build a research-driven outbound package for Tamsin Vale Cemetery."

Process:

• Step 2, The Sourced Profile: 4,200 plots, paper plat maps and one binder per section [CRM]. Two double sales in 2 years [CRM] (illustrative). No owner letter, filing or news found [Unknown]. The signer is [Unknown].
• Step 3, The Zones:

Verdict: MANUAL REVIEW. One zone is HIGH, and no must-have zone is LOW enough to disqualify, but one HIGH zone does not reach GREENLIGHT. Do not write a point of view brief from one strong fact.

Gate: Insufficient Intelligence Fallback. The output has no point of view brief, no signal map and no sequences, so it has 0 emails. No held package is written here either, because the gate stops before Step 7. Held drafts appear only after the research actions close the gap.

Output: What the sources show, and three research actions: find the signer (Zone 3), confirm the pre-need sales (Zone 2), and check what staff use day to day (Zone 4). The signer is the must-have zone with an open fact, so it leads the next action. The rep runs the other two checks in the same pass, because they are the same person's work and fall due on the same default date, 2026-10-26. Recommend gtm-account-snapshot as the faster route to a usable profile.

Hand-off (read by 06):
account: Tamsin Vale Cemetery (TARGET)
run_date: 2026-10-19
pov_brief: none (one HIGH zone is not enough for a point of view)
signal_map: none (three research actions listed instead)
wedge: none (4,200 plots are on file, but the plot price is unknown)
persona_emails: none (0 emails)
verdict: MANUAL REVIEW
zone_scores: 1 HIGH, 2 LOW, 3 MEDIUM, 4 LOW, 5 MEDIUM
account_status: TARGET
report_path: none (inline)
next_action: Nadia Brandt runs gtm-account-snapshot on Tamsin Vale by 2026-10-26 to find the signer, and checks the pre-need sales and the staff tools in the same pass

Example 4: Filing Check Ends In DISQUALIFY (Northvale Memorial Holdings)

Context: A rep asks for a research outbound package on Northvale Memorial Holdings, a public consolidator with about 240 funeral homes and 90 cemeteries. Its annual filing says it runs its own in-house system [Public: annual filing]. The status is OUT, not LOST. OUT means the profile has disqualified the account. A LOST status would mean a closed loss, and there is no loss record here.

Input: "Build a research outbound package for Northvale Memorial Holdings from its annual filing."

Process:

• Step 0 And 1: The user is the rep, who has asked for the package. Status OUT, so the guard does not apply. No AE owns the account.

• Step 2, The Filing (10-K Filing Mode):

  • Public consolidator, about 240 funeral homes and 90 cemeteries [Public: annual filing]
  • Runs its own in-house system [Public: annual filing]
  • Quoted passage (illustrative, not a quote from a real filing): "All of our funeral homes and cemeteries run on one in-house records platform that we build and maintain." [Public: annual filing] (illustrative)

  The passage is an illustrative stand-in. A real run quotes the filing's own words and gives the item and page.

• Step 3, The Zones And The Verdict:

The zones the filing does not speak to are not scored. A blank is unknown, not a no, and the verdict does not need them.

DISQUALIFY. The disqualifiers (profile: Qualification Criteria) include public or PE-backed consolidators that run their own system. This account is one.

• Steps 4 To 7: No signal map, no wedge and no sequences. A disqualified account gets no wedge and 0 emails. The point of the disqualifier is that the company already runs the system PlotIQ would sell into.
• Step 8: Research gaps: none that would change the verdict.

Output: A plain verdict for the rep: DISQUALIFY, with its disqualifier named and its filing source tagged. The rep is told the reason in one sentence, and no outreach is drafted.

Hand-off (read by 06):
account: Northvale Memorial Holdings (OUT)
run_date: 2026-10-19
pov_brief: none (DISQUALIFY)
signal_map: none (DISQUALIFY, the disqualifier is the only signal)
wedge: not computed (DISQUALIFY)
persona_emails: none (DISQUALIFY, 0 emails)
verdict: DISQUALIFY
zone_scores: 1 [Unknown], 2 [Unknown], 3 [Unknown], 4 LOW, 5 [Unknown]
account_status: OUT
report_path: none (inline)
next_action: the requesting rep is told the verdict and its disqualifier by 2026-10-26, and no sequence is drafted

---

Common Patterns

Pattern: 10-K Vs. Earnings Selection

When: A public company has both an annual filing and a recent earnings call. Approach: Use the earnings call for time-sensitive outbound, where the window is 7 to 14 days. Use the filing for deeper strategic outbound, where the relevance lasts months. Both can run on the same company at different times.

Pattern: The Filing Check Before Any Package

When: The source is a public company's filing. Approach: Read the systems section first. A consolidator that runs its own system is disqualified before any zone is scored. That is the cheapest check in the workflow, and it saves the rest of the run.

Pattern: Active Deal Guard (BDR)

When: The user is a BDR and the account has an AE owner who has not said yes. Approach: Write the coordination note for the AE, with the intelligence the AE can use. Write no POV brief and no sequences. The BDR shares the note with the AE rather than sending anything.

Pattern: Insufficient Intelligence Fallback

When: The sources are too thin to score enough zones for a verdict. Approach: Deliver what is available. Mark unscored zones [Unknown], classify the account as MANUAL REVIEW only when at least one zone is scored HIGH, list the research actions by zone, and recommend gtm-account-snapshot as the faster route.

---

Troubleshooting

"The 10-K Content Is Too Long To Process"

Symptom: The filing is too large to read in one pass. Likely Cause: The whole document was pasted. Response: Work the required sections only: Items 1, 1A, 7 and 9A. These carry most of the relevant intelligence. Skip the financial statement tables, and say in the summary that they were skipped.

"The Earnings Call Transcript Is Not Available Yet"

Symptom: The transcript has not been published. Likely Cause: The call is recent, or the company does not publish transcripts. Response: Use the earnings press release as the substitute. It carries the key metrics but not the analyst questions. Tag it [Public: earnings press release, date], and note the reduced depth at the top of the output.

"The Private Company Has Almost No Public Information"

Symptom: Step 2 finds only one or two tagged facts. Likely Cause: The company is small, new, or has no press presence. Response: Classify it as MANUAL REVIEW only if at least one zone is scored HIGH. Otherwise say "Insufficient data" and list the research gaps. Recommend the routes that do exist: a licensing record search, a job posting review, and the company's own staff pages.

"The Revenue Estimation Methods Give Conflicting Results"

Symptom: The employee benchmark and the branch proxy give different figures. Likely Cause: The rates come from different sources, or the company's footprint does not match its headcount. Response: Report the range from each method and name the spread. Tag each figure with its method. Do not average them into a single number.

"A Zone Cannot Be Scored"

Symptom: The sources do not say anything about a zone. Likely Cause: The source is narrow, such as a single filing that covers only financial matters. Response: Mark the zone [Unknown]. Never score it LOW from silence. Name the source that would settle it as a research action.

"The Account Is ACTIVE With An AE Owner"

Symptom: The user is a BDR and the account has an open deal. Likely Cause: The account is in a late stage, and the BDR was not aware of the AE's work. Response: The guard fires. Write the coordination note for the AE and stop. Write no sequences, unless the AE has said yes.

---

Best Practices

Do's

• Tag Every Claim. Use the six tags: [Public: source, date], [CRM], [Call], [Notes], [Inferred: method] and [Unknown]. The tag tells the next reader how much weight the fact can bear.
• Use A Different Anchor In Each Email. A signal used twice in one persona's sequence reads as a copy-paste, and the prospect notices.
• Match Proof Points To The Vertical. Pick the proof point that matches the prospect's segment and persona, and only from Verified or Benchmark status.
• For Earnings Calls, Act Fast. The 7 to 14 day window is real. Speed matters more than polish.

---

Integration With Other Skills

• Reads From gtm-account-snapshot (02), By These Names:
  • account_status: an ACTIVE status with an AE owner triggers the guard, and an ENGAGED status turns the replying persona's sequence into one reply draft, and an INBOUND status turns the requesting persona's sequence into one reply with the other sequences held back until the account goes quiet (Step 1b)
  • verdict: a DISQUALIFY from 02 is confirmed here, not assumed, before research starts
  • contact_priority: the ranked personas and names to write to first
  • pain_hypotheses: each one is tagged. Only a claim tagged [Public: source] or [CRM] may be the claim in an email. An [Inferred] hypothesis is not used as a claim
  • sequence: 02's sequence is not repeated. This skill's sequences add the source-based anchors that 02 did not have
• Account Qualification (01) runs before 02. Its verdict reaches this skill through 02's verdict field, so nothing is read from 01 directly.
• Hands To gtm-meeting-prep (06), By These Names: pov_brief, signal_map, wedge (with its arithmetic shown) and persona_emails. Meeting prep reads them to start the first call from the same facts.
• Hands To Wrap-Up: report_path when a file was written, the verdict and zone_scores as the session's result, and next_action with its owner and date.
• Shared Profile Section: Trigger Event Outbound (04) and Closed-Loss Reactivation (05) read the same Industry Context section. This skill hands them no fields.
• Routes To: gtm-competitive-displacement (12) when a specific incumbent is named in the source, gtm-trigger-event-outbound (04) when the source carries a dated event, and gtm-deal-pulse (08) once an opportunity exists.

Chains

• Qualify To Book: Account Qualification (01), Account Snapshot (02), then Research Outbound (03, this skill) or Trigger Event Outbound (04), then Meeting Prep (06). This skill adds the sourced point of view, the signal map and the wedge for enterprise accounts, where the depth is worth the time.

---

Changelog

2.1.17 (2026-10-27)

• The shared contract now says an ask to the next action's person with no date of its own travels with the next action, and defines the champion test "explains the internal politics" (the contact tells the rep who decides, or how the decision gets made).

2.1.16 (2026-10-27)

• The shared contract's Champion Tests section now defines "commits to next steps with dates": the contact agrees, on record, to a dated next step in the deal, such as a meeting they book or a date they commit to on a call.
• The shared contract's next-action rule now says which asks travel with the next action: every ask to the same person due on or before its date, counts and figures included, and never a document the person must draft first or an ask due later.
• Next action. Step 9 item 3 replaces "with no separate date" and "a date that a record or the skill's own schedule sets" with the default-date test: a check for the same person travels when its date, the shared default, is on or before the next action's date. Example 2 states the default date, 2026-10-26, as the reason its two checks share the pass. No next action changes.

2.1.15 (2026-10-27)

• The shared contract's Champion Tests section now defines "asks how to help" (the contact asks what the rep can supply to move the internal decision). The PlotIQ profile dates Brackhaven's deal from 2026-05-06, names the economic buyer for the Stage 2 gate the way the Buying Committee does, and limits the $25K board line claim to one-site accounts.

2.1.14 (2026-10-27)

• The shared contract's Champion Tests section now defines "creates urgency" (a dated deadline the decision works to, with its date and source on record) and "defends the solution" (a recorded case of arguing for us in our absence to someone who decides), and says a test not yet run is never scored.

2.1.13 (2026-10-26)

• The owner rule (Step 9, Owner Naming) says a person the profile's Example Cast names for an internal role is on record for that role, and the owner is written with the citation. Step 7 points to it. The Sales Manager clearance in Example 1 and the try-it names Lena Vasquez with the citation.
• The cost basis line (Step 5) states the recurring year and the one-time setup as two figures, and does not sum them. Example 1 already presented them this way, so its figures are unchanged.
• An expired trigger window does not change the routing (Step 7). The summary states the signal's age in days beside its window. Example 1 states the owner letter as 231 days old.
• The records state the last system decision as an event, seven months from first meeting to signed contract, and not in the words of the Zone 3 LOW cell. The Zone 4 record reads as a fact about the plot lists. Zone 3 and Zone 4 scores and the MANUAL REVIEW verdict are unchanged. The email that cites the cycle cites the same seven months.

2.1.12 (2026-10-26)

• The next action has one priority order, written in the Output Contract next_action row and in Step 9. A reply owed to a buyer comes first, then AE coordination for an ACTIVE account with an AE owner, then the verdict's step. Ties go to the earlier buyer message or due date, then to the lower Buyer Personas number.
• A MANUAL REVIEW hold is cleared by the Sales Manager, the role the profile's Example Cast names. The BDR who drafted the package does not clear its own hold. The clearance is the next action, and the first send is a later step (Steps 3 and 7, Hard Rule 11, the Quality Bar).
• Asks to the same person with no separate date travel with the next action. A step for a different person, or with a date set by a record or the skill's own schedule, is a later step.
• After the trigger-named personas, the rest follow in Buyer Personas table order, 1 to 5.
• The Operations Director applies to an account with a site that keeps plot records or runs daily work orders, with or without a named contact.
• The wedge has a benefit line and a cost basis line. Its Benchmark is the profile row whose record type matches the account's, and a Benchmark with a different record type is not used.
• Step 9 item 6 and the cheatsheet state the line "Region tier: no tier yet" when the state is not on record.
• Open items are the Step 8 gaps. A gap caused by a missing field names the field and is not stated as a result.
• Example 1 Step 3 says "the reasons are" where it said "the open items are". Its cheatsheet heading matches the Cheatsheet Layout name, Geographic Or Market Exposure.
• Re-derived next actions: Example 1 (Saltmarsh) is the Sales Manager's clearance, with the send as a later step. Example 2 (Brackhaven) moves the AE's decision to a later step. Example 3 (Tamsin Vale) bundles its two same-rep checks into the next action. Example 4 (Northvale) names the requesting rep.
• The try-it promise changes its next action, its hold line, its cost basis label, its region tier line and its open item note. No verdict, zone, count or figure changes.

2.1.11 (2026-10-26)

• Step 7 states that a persona applies with or without a named contact. A persona with no name is written by title, and its gap is named in the summary.
• Step 5 states the wedge unit. A profile price counts one site, cemetery or funeral home alike, because the pre-need module serves funeral homes. The count is five sites in the Saltmarsh example, and two cemetery sites is not the unit.
• Step 5 states the payback unit for a plot-count benefit: one unsold plot, priced at its average sale value, which the account or the profile must source.
• Step 4 names the one source that states a signal when two sources hold it.
• Step 3 states that a must-have zone at LOW keeps the verdict out of GREENLIGHT, and the reasons name it.
• The records state the decision cycle as a fact. The mode comes from the source type, so the records no longer state it, and the Saltmarsh facts stay the same.
• No try-it promise, verdict, zone score, wedge figure or email count changes.

2.1.10 (2026-10-26)

• Gone quiet is no longer restated locally. Step 1b points to the shared rule in What The Close Reads.
• An INBOUND account with a DISQUALIFY verdict gets one short decline draft to the requesting contact, held for the rep to read, in Steps 3 and 7, the Output Contract, Hard Rules 6 and 14 and the Quality Bar.
• No example is an INBOUND account, and no try-it promise changes.

2.1.9 (2026-10-26)

• INBOUND accounts, which asked us first with no open deal, get one reply to their request for the requesting persona, carrying this skill's angle. The other persona sequences wait until the account goes quiet, meaning no reply within five business days after the reply is sent.
• The INBOUND rule sits in Step 1b, Step 7, the Output Contract, the Integration section, Hard Rule 13 and the Quality Bar.
• No example changed, and no try-it promise changed.

2.1.8 (2026-10-26)

• Every profile citation is written (profile: Section), naming the real section. Template placeholders and the shared contract are unchanged.
• A MANUAL REVIEW sequence is a held package: its first line is marked Held, nothing in it goes out before a person clears the verdict, and Hard Rule 11 says so. Account Snapshot (02) holds its MANUAL REVIEW sets the same way.
• ENGAGED is one of six statuses. Step 1b gives an ENGAGED account one reply draft to its open thread for the replying persona, not a cold sequence, and Hard Rule 12, the Quality Bar and the Integration section follow it.
• The trigger leads every persona's sequence. The first line of each email names the trigger source, and every one of Example 1's twenty emails now opens on the owner letter.
• Example 1 restores the Saltmarsh decision-cycle fact, in the words the records hold: the group reviews system spend twice a year, and the decision cycle is 6 to 9 months [CRM], illustrative. Saltmarsh stays Zone 3 LOW and MANUAL REVIEW, and its 20 emails are held.
• Example 3 stays at 0 emails. Its fallback gate stops before Step 7, so there is no sequence to hold until the research actions close the gap.
• Headings and labels are capitalized word by word.

2.1.7 (2026-10-26)

• The count rule is one rule. Every applicable persona gets a sequence, and the persona count comes from the profile and the account. The trigger row sets the order and the lead, and it does not remove a persona. The rule text no longer carries a worked count.
• The business-it-serves test is defined: a persona applies when the account runs the business that persona owns.
• A profile price is per site. The unit is one site in the account's footprint, cemetery or funeral home alike.
• A Benchmark about plot records applies to a site whose plot list is kept by hand, and the wedge names the record type.
• A MANUAL REVIEW account gets its wedge and sequences. Account Snapshot (02) writes sequences for GREENLIGHT only, and this skill says so.
• The rule text no longer refers to the examples by number. The try-it pointer names a full first run.

2.1.6 (2026-10-26)

• Example 3 re-scores its zones on the rules and the records. Zone 2 LOW cites the recorded no-pre-need fact, not a website check. Zone 3 MEDIUM cites the recorded approval path and the next board meeting. Zone 4 LOW cites a named web search that found no website, as a checked absence. Zone 5 MEDIUM cites the recorded growth project. The zone line and the MANUAL REVIEW verdict are unchanged.
• Example 2 tags the Stonemark evaluation to its dated call.

2.1.5 (2026-10-26)

• The Zone 3 sentence of Example 2 quotes the profile's Zone 3 cell as written.
• The next action follows the family's one deadline rule (five business days after the run date, or an earlier date the records or the skill's own schedule set for that step). The Output Contract and Step 9 point to that rule, and each example next action carries its date under it.

2.1.4 (2026-10-26)

• Try-it evidence completed so a strict first run reproduces Example 1.
• The Example 1 owner letter gains its day, and the try-it records gain the same day.
• Step 5 states the Benchmark rule for a figure stated per cemetery, and a payback rule: a payback figure needs a sourced unit price.
• Step 7 states the persona rule: a persona applies when the account has the business it serves or when the trigger row names it, and the approver is read from the records.
• The next_action in the Output Contract, Step 9 and the hand-off gains a date rule and a scope rule: the review of the package and the first send.
• Step 9 lists the Step 8 research gaps as the open items.

2.1.3 (2026-10-26)

• Try-it evidence completed so a strict first run reproduces Example 1.
• Records gain the company's pre-need intake form and social-ads facts (both illustrative), so Zone 2 meets all three HIGH conditions, and the group's twice-yearly budget review, which sets the 6 to 9 month decision cycle from the group's own process. The owner-letter quote is marked as the wording these records hold.
• Step 3 gains the partial-match sentence, reused from Account Qualification (01) Step 4 word for word.
• Step 5 gains one sentence: a Benchmark stated without a unit is read per cemetery site and multiplied, with the multiplication shown.
• Step 7 gains one sentence: when a trigger row names two personas, the one who approves budget leads.
• Example 1 gains the five-part POV brief, the persona emails (each opening with a citation), the eight-section cheatsheet, the eight-item intelligence appendix and the summary.

2.1.2 (2026-10-25)

• Try-it evidence completed so a strict first run reproduces Example 1.

2.1.1 (2026-10-25)

• Adds examples/try-it.md so a first run reproduces Example 1. No other change.

2.1.0 (2026-10-25)

• Reads the Context OS account folder and offers read-only CRM and call connectors through the shared contract's new Knowledge Base and Connectors section. Adds a Chains subsection. No other change.

2.0.1 (2026-10-20)

• Every hand-off block carries a run_date line after its account line, so two instruments on one deal are compared by date (found by the wave two hand-off tests). No other change.

Version 2.0.0 (2026-10-20)

• Industry Context is read from the profile as a sourced list. An unsourced industry claim is dropped.
• The persona count comes from Buyer Personas, and the email count is personas times steps. The fixed count is removed.
• Step 7 reads the Persona who acts column of the Buying Triggers table.
• The six evidence tags replace the old labels. The Verified, Estimated and product-benchmark labels are retired.
• Zones and the verdict come from the profile. A zone the sources do not speak to is [Unknown], not LOW.
• Proof points are named only when Verified or Benchmark.
• New Example 4: a filing check that ends in DISQUALIFY with zero emails. Example 3 now shows all five zones. Every example ends in a hand-off block.
• Structure brought to the house format: Why This Skill Exists, Role, Modes, Epistemic Rules, Hard Rules and Quality Bar added. The old file-format options are replaced by one file option in the Output Contract.
• Hand-off fields named in the Output Contract and in Integration. The report is inline, and a Markdown file is written only when asked.

Version 1.3.0 (2026-10-13)

• Examples re-cast on PlotIQ, the library's one example company. Example errors fixed and a non-success example added.
• Persona count corrected to 5, giving 20 emails per 10-K or private package.

Version 1.2.0 (2026-09-29)

• Worked examples rewritten around one shared case.
• Em dashes removed from prose.

Version 1.1.0 (2026-07-06)

• Restructured around the five-part skill anatomy: Role, Input Contract, Output Contract, Context, Methodology.
• Company-specific data moved out of the skill and into the shared profile.

Version 1.0.0 (2026-03-04)

• Initial release, merged from three research workflows (10-K, private company, earnings call) and unified by research type.