agentleFS
getcargohqgetcargohq/gtm-skills/tam-building-advanced/SKILL.md

tam-building-advanced

Skill2 starsChanged 29 days agoMITView on GitHub

Build your total addressable market (TAM) from several sources as one account layer in Cargo: one brief goes to Sales Navigator, AI Ark lookalikes, Exa and Apollo, Cargo merges every company into one account, each is classified into your own categories, flagged in or out of scope with the reason and marked net new or already in your CRM, and the report estimates what no source found. Triggers: "build our TAM from several data providers", "how much of our market do we actually have", "which of our TAM is net new and which is already in the CRM", "estimate the size of our market from our own data". Cargo project, salesNavigator, aiArk, exa, linkedin, unifyAccounts, classify. Skip when: one source and one model are enough, which is tam-building; a list in the chat or a CSV, which is build-tam-list; or only a Sales Navigator search past its cap, which is extract-sales-navigator-search.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 4 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    One account per company, merged by Cargo, not by a workflow: Every source lands its rows in its own model and declares how it identifies a …

    line 46
  2. Distinctive:

    Your categories, not the providers': Every company is classified from its own description into the categories you set, ideally the values …

    line 56
  3. Very specific:

    Every substantive message starts with the current phase and ends with a Next step section giving the operator one concrete decision, what …

    line 103
  4. Very specific:

    ICP and categories: Read the context for an ICP.

    line 107
  5. Very specific:

    Install the required authoring skill first: If cargo-project is absent, run:

    line 130
  6. Very specific:

    Reconcile it with what is already declared: If the project already has any of the five connectors, an accounts declaration (Unify Accounts, …

    line 147
  7. Very specific:

    Source and report: Follow references/run.md: one source at a time, the profile batch piloted on five accounts before the rest, then the …

    line 168
  8. Very specific:

    every Sales Navigator URL is under 1,000 accounts and carries INDUSTRY, COMPANY_HEADCOUNT and REGION

    line 264
  9. Distinctive:

    every account a TAM source found has one profile with a LinkedIn company id and a status, no LinkedIn company id sits on two accounts, and …

    line 267
  10. Distinctive:

    Read this before pointing the skill at a market-sized brief.

    line 276

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

TAM building, advanced

The outcome

Your TAM as one account layer in Cargo, built the way no single data provider can: the same brief given to four sources, every company they find merged into one account by Cargo itself, and a report that says how much of the market you hold, how much of it is already in your CRM, and how much no source found.

tam-building is the single-source version (one AI Ark search, one model). This one is for the question that comes after: how much of my market do I actually have?

Four parts, each with its own way of going wrong.

One brief, four sources. Sales Navigator is the census (every company LinkedIn files under your filters, split under its 1,000-account cap). AI Ark finds lookalikes of the customers you name. Apollo is a second census in a different database. Exa searches by meaning, which reaches companies filed under industries nobody filters on. Industries, company size and countries are written into all four, so they search the same market. A source missing one of them bills for the whole dimension.

One account per company, merged by Cargo, not by a workflow. Every source lands its rows in its own model and declares how it identifies a company: Sales Navigator by the LinkedIn company id, AI Ark and Apollo by the domain and the LinkedIn page, Exa by the domain. Cargo's account unification merges them into Unify Accounts and normalises the keys itself. Then one play reads each account once: a LinkedIn company enrich (from the LinkedIn page any source gave, else the domain) and a classification, written as one profile row in tam_accounts. That row carries the LinkedIn id, the page, LinkedIn's domain and the account's own domain together and unifies back into the same account: it is what joins the company Sales Navigator found to the same company AI Ark found. A company four sources found is enriched once.

Your categories, not the providers'. Every company is classified from its own description into the categories you set, ideally the values of your CRM's industry field. A provider's industry label is never trusted: LinkedIn files plenty of SaaS companies under IT services, and plenty of agencies under software. Nothing is deleted: a company outside the ICP stays with status and exclusion_reason, so you can see what each source returned that was not your market.

Net new, known, and missing. Your CRM's companies already unify into the same accounts, so "already in the CRM" costs nothing: it is an account that also holds a CRM row. Every account also says which sources found it. The report then estimates the companies no source found from the Sales Navigator × Apollo overlap, and gives your coverage as a floor.

Example

Build our TAM from Sales Navigator, AI Ark, Apollo and Exa for B2B software with 51 to 1,000 people in North America and Western Europe, and tell me how much is net new.

Illustrative output, fictional records:

TAMCompanies
In scope18,420
Already in your CRM4,960
Net new13,460
Flagged out of scope (with reason)21,305
Needs review (thin page)1,140
SourceFoundFound by no other sourceAdded in this order
Sales Navigator16,2109,38016,210
AI Ark5,640910+1,250
Apollo7,980640+730
Exa1,520120+230

Coverage: Sales Navigator × Apollo puts the market at 22,100 or more, so the 18,420 in scope are about 83% of it at most. Top account: northwind-analytics.example, B2B SaaS, 340 people, found by all four sources, net new.

Guide the operator through every phase

flowchart LR
  icp["1. ICP and categories"] -->|"Approve brief, categories"| sources["2. Sources"]
  sources -->|"Approve filters, limits, spend"| build["3. Build"]
  build --> run["4. Source, one at a time"]
  run --> report["5. Report"]
  report --> next["Openings: score, push net new, enrich, review gaps"]

Every substantive message starts with the current phase and ends with a Next step section giving the operator one concrete decision, what it unlocks, and what stays blocked. During an in-progress operation, say No action needed and name the next checkpoint. Never end with a generic offer to help.

  1. ICP and categories. Read the context for an ICP. Settle industries, company size and countries, each quoted from icp.md or labelled a suggestion. Derive the categories from the CRM's company industry field, or from icp.md. Ask the operator to approve the brief and the categories.
  2. Sources. Translate the brief into each source (references/configure.md): Sales Navigator searches (split with extract-sales-navigator-search past 1,000 accounts), AI Ark groups and lookalike seeds counted for free, Apollo with a small first limit, Exa queries. Present each pool, each limit, the live prices and the estimate, including one LinkedIn enrich and one classification per distinct company. Ask for one approval of the filters, the limits and the maximum spend.
  3. Build. Adapt, run node --import tsx evals/contract.mjs, check, plan, deploy.
  4. Source, one at a time. Sales Navigator, then AI Ark, then Exa, then Apollo. Sync a source, let Accounts merge it, run the profile play once as a batch over its new accounts (five first, then the rest), then the next (references/run.md). The play is a build step: no schedule, never enabled.
  5. Report. The headline (TAM, net new, already in the CRM), what each source added, what each returned that was not your market, the coverage estimate, actual credits against the estimate, and the openings.

Put it in your project

This folder is a worked example: real project resources written for some other company. The job is to end up with the code your company would have written, in your project, and an agent does the adapting.

Install the required authoring skill first. If cargo-project is absent, run:

npx skills add getcargohq/cargo-skills --skill cargo-project

Then read .agents/skills/cargo-project/SKILL.md directly; no session reload is needed. Complete its bootstrap and use its authoring, state, plan, and deployment rules throughout. Stop before any research or template work if the skill cannot be installed or read.

  1. Install it: the CLI does the copy. From inside the project (the folder holding cargo.json), cargo-ai project add cookbook/tam-building-advanced writes this example to infra/tam-building-advanced/ and this procedure to .agents/skills/tam-building-advanced/. No project yet? cargo-ai project init <dir> --cookbook tam-building-advanced && cd <dir> && npm install does both; this folder never ships a shell. If you are reading this from the project's .claude/skills/, the install already happened: start at step 2. On a CLI too old to have add, copy this folder in as a sibling of what is there by hand; everything below is unchanged.
  2. Reconcile it with what is already declared. If the project already has any of the five connectors, an accounts declaration (Unify Accounts, from readiness-scoring or account-scoring) or these folders, rewire the imports to the existing ones and drop the copies: two resources with one slug collide at deploy. Check that the workspace has Unify Accounts (cargo-ai storage model list, extractor unifyAccounts) and that the CRM's companies model unifies into it (references/configure.md). This skill declares no defineContext. This folder needs nothing in .env; append nothing and never overwrite it.
  3. Write or confirm the ICP, the three criteria and the categories. Follow references/configure.md sections 1 and 2, and stop for approval.
  4. Translate the brief into the four sources and size them. Same file, sections 3 and 4. Stop for approval of the filters, the four limits and the spend. Do not deploy, sync or bill anything while that approval is pending.
  5. Adapt and deploy. Work the sections below in order: What should not change is what you argue back about (say what breaks, then do it if they still want it); What you can change is what you offer unprompted (nobody asks for a variant they do not know exists); What you will be asked is the floor, and you derive before you ask. If you are asking more than about four questions you have skipped lookups. Record what you changed and why under a ## Decisions section in your copy of this file. Then run node --import tsx evals/contract.mjs from this skill's folder, followed by cargo-ai project types && cargo-ai project check && cargo-ai project plan from the project root. Show the diff and deploy. Never run cargo-ai project init --force in a non-empty directory.
  6. Source and report. Follow references/run.md: one source at a time, the profile batch piloted on five accounts before the rest, then the report queries, and walk Done when line by line with evidence. Deployed cleanly and produced nothing is the normal failure.

What you will be asked

Derive before you ask. An input with a lookup is looked up, not asked. Only the approvals genuinely live in the operator's head.

InputKindHow it is answeredWhy it matters
icpderivedThe workspace context first (icp.md, a scoring rubric). None? tam-building's research step, after its prices are approvedFour source filters and the classifier answer to it. If they answer to four different readings, the TAM is four markets
market_minimumderivedIndustries, company size and countries, quoted from icp.md or labelled a suggestion, confirmed once and written into all four sources and settings.tsA source missing one bills for the whole dimension, and the sources stop searching the same market
categoriesderivedThe CRM's company industry field when a CRM is connected (read the property's options), else the categories section of icp.mdThey are what every company is classified into. Your CRM's values make the TAM and the CRM speak one language
sales_navigator_searchesderivedBuild the search from the three criteria and count it. Over 1,000 accounts: extract-sales-navigator-search splits it and returns the URLsOne search serves at most 1,000 accounts; past that the rest of the market is silently not there
lookalike_seedsderivedUp to five customers: the ones the context names, or the CRM's closed-won companiesAI Ark's lookalikes are only as good as the seeds; a guessed customer is not evidence
exa_queriesderivedOne per category and country, in the words a buyer would use, in exaQueries (settings.ts)Exa reaches companies filed under labels nobody filters on, but only for what the queries describe
crmderivedThe CRM companies model that unifies into Unify Accounts (cargo-ai storage model list, then its key in an account's ids). None: every company is net newIt decides what "net new" is measured against
limitsderivedFrom each source's count or a small sample pull, as a fraction of the pool for the first runEvery paid source bills per returned record: the limits are the invoice

The two approvals (the brief with its categories, then filters plus limits plus spend) are decisions, not inputs: they are asked every time.

Checked before moving on, not after the deploy:

  • market_minimum: the same three criteria appear in every source's config and in settings.ts
  • categories: the in-scope list, the out-of-scope list and "Not enough information" are all present
  • sales_navigator_searches: every URL is under 1,000 accounts and carries INDUSTRY, COMPANY_HEADCOUNT and REGION
  • limits: each one was shown against its pool with a live price

What you can change

The code is a worked example. These reshapes are expected, and the agent offers them rather than waiting to be asked. Every one costs something; that is what makes it a variation and not the default.

VariationWhen it is rightHowWhat it costs
apollo-as-fallbackBudget: three sources are enough until the report shows a gapLeave tam_source_apollo out of the first deploy and add it when the per-source table says Sales Navigator, AI Ark and Exa missed accounts you expectedNo Sales Navigator × Apollo pair, so no coverage estimate unless a reference-list replaces it
reference-listYou hold an independent list of the same market: a Crunchbase export, an earlier TAM, a list built another wayLoad it as a model (Google Sheets or CSV) that unifies as an account on its domain or LinkedIn page, and pair it with Sales Navigator in the coverage query instead of ApolloOne more source to profile. The estimate gets better the more independent the list is; a list built from the same providers adds nothing
drop-a-sourceBudget, or a source that adds almost nothing in your marketRemove its connector, its model and its line in the play's filter, and its entries in the contract, in one changeCoverage. Dropping Apollo also removes the census pair the estimate is computed from; pair Sales Navigator with AI Ark and say the estimate is weaker
growth-signalGrowth decides priority in your marketAdd salesNavigator.findCompanyInsights after the LinkedIn enrich in the profile play, and a growth column on tam_accountsOne more paid call per company. Keep growth as information, never as a gate: a company that is not growing is still in your market
refreshThe market moves and the TAM goes stale. Every three to four months is enoughBy hand: re-count and re-sync the sources in the same order, and run the profile batch after each. Or on its own: schedule: { type: "cron", cron: "0 6 1 */3 *" } on each paid source (the 1st of every third month), and "0 6 2 */3 *" with isEnabled: true on the profile play, the day after. A new company becomes a new account and is profiled; a known one is merged and costs nothing moreEvery re-sync re-bills every returned record, including the companies already in the TAM: four times a year at most, and the contract refuses anything more frequent. Re-count Sales Navigator first: a split search drifts over a quarter
fifth-sourceA source that is good at your market (FullEnrich, free per record)One more source model with a declared account unification, one more line in the play's filter, and the contract's connector list editedNothing else: Cargo merges it, and the play profiles only the companies it adds

What should not change

However far you adapt, these hold. Ask for one anyway and the agent tells you what breaks, then does it if you still want it, and records why under ## Decisions in your copy of this file.

  • Cargo merges; no workflow writes the TAM. (unification on every model in infra/models/) Unify Accounts is the TAM. A play that dedups sources into its own table re-implements key normalisation, misses the merges Cargo makes across three rows, and leaves two account layers that disagree.
  • Every unification is declared. Sales Navigator's default unification is off: leave it undeclared and the census never reaches Accounts. The others are declared too, so a provider changing its default cannot move your TAM.
  • One profile per company, after the merge. (infra/plays/profile-company.ts) The LinkedIn enrich and the classification run on the account, so a company four sources found is paid for once, and the profile row is what joins the Sales Navigator row to the others.
  • The profile matches on both domains. (infra/models/tam-accounts.ts) LinkedIn's domain and the account's own often differ (corporate.acme.example against acme.example). Match on LinkedIn's only and the profile becomes an account of its own while the source account stays unprofiled.
  • A company's domain comes from evidence, never from a short link. (infra/scripts/company-domain.ts) LinkedIn's domain when it is real, else the account's, else where LinkedIn's short link lands. A bit.ly or hubs.ly host written as a domain is a key shared by unrelated companies, and Cargo merges them into one account.
  • Companies are classified from their own description into your categories. (infra/plays/profile-company.ts, infra/settings.ts) Trust a provider's industry label and services firms land in the TAM while real software companies filed elsewhere are dropped.
  • Nothing is deleted; everything out of scope carries its reason. (status and exclusion_reason, computed on tam_accounts) Delete instead and the report can no longer say what each source returned that was not your market, or why a company you expected is missing.
  • Nothing else is declared on Unify Accounts. (infra/models/accounts.ts) A declared column list or schedule replaces the live one, and every pipeline and person in the workspace reads that model.
  • Unknown stays unknown. A thin page, or a missing headcount or country, is needs_review, never out of scope. A provider that returns nothing has a coverage gap; it is not evidence about the company.
  • Sources run one after another. (references/run.md) Profile one source's accounts before the next syncs: the next source's rows then merge onto accounts that already have a profile, and only the companies it adds are paid for.
  • The three criteria are in every source. (infra/models/tam-source-*.ts) A source without industries, size or countries bills for the whole dimension and searches a different market than the others.
  • A paid source is built once, refreshed at most quarterly. A sync re-bills every returned record, including the companies already in the TAM. No schedule by default; the refresh variation's quarterly cron is the most frequent the contract accepts.
  • Sales Navigator runs without a seat. No identityIds: a search through someone's own account puts that account at risk.
  • Nothing is written to the CRM. Pushing the net new is a separate write the operator approves, with a reviewed sample first.

Done when

  • the ICP came from the context or was researched and approved, and the three criteria are in all four sources and settings.ts
  • the categories came from the CRM's industry field or icp.md, and every row of tam_accounts carries one of them
  • every Sales Navigator URL is under 1,000 accounts and carries INDUSTRY, COMPANY_HEADCOUNT and REGION
  • the sources were synced and profiled one at a time, in order, each piloted on five accounts first
  • every account a TAM source found has one profile with a LinkedIn company id and a status, no LinkedIn company id sits on two accounts, and every account holding two LinkedIn ids was read
  • the report gives the TAM, net new and already in the CRM, the per-source table, the coverage estimate (or why it is too thin), and actual credits against the estimate
  • nothing was written to the CRM
  • node --import tsx evals/contract.mjs passes against the adapted resources

What it costs

Read this before pointing the skill at a market-sized brief.

Counting is free on AI Ark and a fixed fee per search on Sales Navigator. Count every candidate before anything syncs.

Sourcing bills per returned record on Sales Navigator, AI Ark and Apollo, and per search plus per result past the tenth on Exa. Each source's limit (or the number of queries times exaResultsPerQuery) is its cap.

Enrichment bills once per company, not per source. Cargo merges the sources before the play runs, so a company four sources found is one account and one profile: one LinkedIn enrich and one classification. Estimate them on the expected number of distinct companies, which is well under the sum of the sources. The merge itself is free.

Fetch every live price immediately before the approval with cargo-ai connection integration get for salesNavigator, aiArk, apolloio, exa and linkedin, and record the lookup time. No price is written in this skill on purpose: they change.

The sources carry no schedule. See refresh above for what a re-sync really costs.

Composes into

account-scoring tiers the in-scope companies of tam_accounts against the same ICP; compare the tiers by number of sources, the companies several sources agree on tend to score higher. Pushing the net new into the CRM is the usual next write, approved on its own; crm-enrichment and crm-deduplication keep those records clean once they are there. extract-sales-navigator-search is the step that splits a large Sales Navigator search, and tam-building is the single-source version of this skill.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.