richapiai/gtm-skills

34 skill entries

• account-research

  Everything knowable about one account, assembled into a brief a rep can act on — firmographics, the site's own evidence, the buying committee, what the company is saying, and where its money and traffic come from. One cheap default pass; the deep passes are opted into and priced first. Use when asked to "research Acme", "tell me about <company>", "who should I talk to at X", "account mapping", "brief me before this call", or "what's going on at <company>". Proactively suggest before any outreach step: copy with no research behind it is a guess with a send button. (richapi-gtm)

  Self-containedDeclared tools4,513 words
• ads-audience

  Turns a list the pack already holds into a matched audience an ad platform will actually accept — suppression-filtered, hashed, and checked against the platform's own minimum BEFORE a credit is spent, because an upload under the floor is rejected after the money is gone. Use when asked to "build a LinkedIn matched audience", "make a Meta custom audience", "Google Customer Match list", "retarget this list", "upload these contacts to ads", or "why did the platform reject my audience". Refuses on an unknown platform, refuses below the floor, and never writes a sender export. (richapi-gtm)

  Self-containedDeclared tools3,478 words
• build-prospect-list

  Turns a plain-English ICP (or a target account list) into a deduplicated prospect list, routed to the cheapest search that can express the filters, page-gated, and costed from the generated catalog before anything is spent. Use when asked to "build a list", "find 40 VPs of Sales", "source prospects", "who works at these accounts", "people who just changed jobs", "who commented on this post", or when a Sales Navigator URL or a LinkedIn post URL is pasted. Proactively suggest before /enrich-waterfall when the user has no list yet. (richapi-gtm)

  Self-containedDeclared tools4,294 words
• call-intel

  Turns a call transcript into structured intel a team can act on (objections raised, next steps agreed, competitors named, commitments made), with every item anchored to a verbatim line of the transcript, and an explicit null wherever the call produced nothing. Use when asked to "summarise this call", "what were the objections", "pull the next steps out of this transcript", "who did they mention", "what did we commit to", or "log this call". A call with no objection returns no objections; it never returns a plausible one. (richapi-gtm)

  Self-containedDeclared tools3,193 words
• campaign-review

  Reviews a contact list against the pack's quality floors AND against /comply's verdict for the same list, then emits a PASS or FAIL bound to the list's content hash. A list with no live compliance clearance FAILs here. Use when asked to "review this campaign", "is this list ready to send", "check this list before launch", "sign off on this list", or "why did launch refuse". Makes zero API calls and spends nothing. Proactively invoke before any hand-off to a sending tool: /launch will not write an export without a verdict this skill produced. Run /comply first. (richapi-gtm)

  Self-containedDeclared tools3,778 words
• competitive-intel

  What a competitor is actually doing (the ads they are buying, the stack and pixels on their site, what they publish, and where their traffic and demand come from), assembled as evidence with an endpoint named against every line. One cheap flat default tier; the deep tiers are opted into and priced first. Use when asked "what ads is X running", "compare us to Y", "what tech does X use", "competitive scan", "battlecard", "are they outspending us", or "what is X posting about". Proactively suggest before positioning work and before a competitive deal review. (richapi-gtm)

  Self-containedDeclared tools4,213 words
• comply

  The hard compliance gate. Resolves the jurisdiction of every contact, clears a lawful basis under GDPR, CCPA/CPRA or CASL before anything is contacted, runs the fail-closed suppression check, enforces retention, and executes erasure requests. Writes a compliance verdict bound to the list's content hash; /campaign-review reads it and FAILs without a live PASS, so /launch cannot export a list this skill stopped. Use when asked "is this list safe to send", "can I email this contact", "delete this person's data", "right to be forgotten", "GDPR", "CCPA", "CASL", "do not contact", or before any launch. Proactively invoke before any review or export step — this skill can stop a run, and a run it did not clear is not cleared. (richapi-gtm)

  Self-containedDeclared tools10,484 words
• cost-optimizer

  Finds where the credits went and what to do about it, from the real ledger. Use when asked "where did my credits go", "why is this so expensive", "how do I spend less", "what is wasting credits", or "audit my spend". Reports savings as ranges bounded by what the ledger can prove, names the evidence under every recommendation, and says when that evidence is an estimate. Makes zero API calls and spends nothing. (richapi-gtm)

  Self-containedDeclared tools5,027 words
• crm-export

  Writes the pack's data out as a CRM import file (HubSpot, Salesforce, Pipedrive, Close, Attio, or a canonical generic file), suppression-filtered at write time, with a manifest saying exactly what PII is in it. Use when asked to "export this to HubSpot", "get this into Salesforce", "CRM-ready file", "import file for my CRM", or "export as CSV for my CRM". It is NOT the sender export: any target the pack already treats as a sending tool is refused by name and routed to /launch. Makes zero API calls and spends nothing. (richapi-gtm)

  Self-containedDeclared tools5,579 words
• crm-sync-expert

  Designs a CRM synchronisation before anyone runs one — which object a row belongs to, which key deduplicates it, what wins on a collision, and what the CRM will silently truncate, coerce or drop on import. Use when asked "sync this to HubSpot", "push to Salesforce", "map these fields", "which dedupe key", "two-way sync", "custom fields for enriched data", "lifecycle stage", or "why did my import create duplicates". It advises and prepares; it never writes to a CRM, because no endpoint in this pack can. (richapi-gtm)

  Self-containedDeclared tools3,016 words
• enrich-waterfall

  Waterfall enrichment over a contact list (profile, work email, phone, verification), with the cost shown and approved before anything is spent. Use when asked to "enrich this list", "find emails for these people", "get phone numbers", "fill in the missing fields", or "why is my coverage so low". Proactively suggest after any list-building step: an unenriched list is a dead list. (richapi-gtm)

  Self-containedDeclared tools2,663 words
• evidence-score

  Grades how well-evidenced a claim about an account or person actually is, and rolls those grades into an auditable 0-100 readiness score across fit, timing, influence, engagement and reachability. Every point traces to a signal with a source line; a dimension with no signal scores zero and says so. Use when asked to "score these leads", "rank by signal", "who should I contact first", "prioritise this list", "is this claim supported", or "can I say this in an email". Proactively invoke before /personalize — copy may only assert what this skill graded supported. (richapi-gtm)

  Self-containedDeclared tools3,151 words
• gtm-kickoff

  The entry point for a new GTM engagement. Interrogates the motion before a single credit is spent (who the ICP really is, whether demand exists, what the wedge is, which channels are hypotheses rather than plans), challenges the premise, and writes a dated strategy brief that every later skill reads. Use when asked "help me build a GTM motion", "where do we start", "we're launching X", "who should we sell to", or when any GTM request arrives with no ICP and no list. Proactively invoke before /icp-review or /build-prospect-list when nothing has been established yet. (richapi-gtm)

  Self-containedDeclared tools2,522 words
• gtm-onboard

  Teaches the pack who YOU are. Captures the seller (company, website, what you sell, the wedge, what you may cite, how outreach should read, who owns which inbound) into `gtm/profile.yaml`, and records standing dos and don'ts into `gtm/preferences.jsonl` so the next session starts knowing them. Optionally pre-fills the profile from your own website, priced and declinable like every other paid call. Use when asked "set me up", "onboard me", "remember this", "never do X again", "always do Y", or when any skill needs the offer and nothing on disk states it. Proactively invoke on a first run, before /gtm-kickoff. (richapi-gtm)

  Self-containedDeclared tools1,661 words
• gtm-retro

  A retrospective across many runs and campaign arms, ending in decisions: stop this play, keep that one, scale the third. Groups runs by the journal's hashed list key, compares arms on coverage and on credits-per-outcome as ranges, and refuses to declare a winner whose range overlaps the loser's. Use when asked "what should we stop doing", "which campaign is working", "retro on last quarter", "compare these two plays", or "where is the budget going". Reads the ledger, the run journals and the activation store; makes zero API calls and spends nothing. (richapi-gtm)

  Self-containedDeclared tools5,507 words
• icp-review

  Tests an ICP against accounts that actually closed, and writes the anchor artifact every downstream skill reads. Samples won and lost accounts, measures each proposed attribute against both, refuses any criterion that cannot name its evidence, marks negative personas and the anti-ICP, and versions the result to gtm/icp.yaml. Use when asked "who is our ICP", "validate our ICP", "which accounts should we target", "why are we losing these deals", "tier our accounts", or "is this segment worth it". Proactively invoke after /gtm-kickoff and before any list is built. (richapi-gtm)

  Self-containedDeclared tools2,881 words
• inbound

  Routes one inbound lead (a demo request, a trial signup, a contact form) to the right owner in under five minutes, on a fixed, flat-priced recipe that was named and costed before the lead ever arrived. Enrich one person, enrich one company, decide, hand off. Use when asked to "route this lead", "who owns this demo request", "someone filled in the form", "qualify this signup", "is this lead worth a call", or "triage inbound". Runs unattended only under a recorded standing approval; with no approval and no human, it queues rather than guesses. (richapi-gtm)

  Self-containedDeclared tools3,583 words
• launch

  Writes the sender-format export (the last artifact the pack controls) and only against a PASS verdict bound to the list's content hash. Use when asked to "launch this campaign", "export for Smartlead", "export for Instantly", "hand this list to the sender", or "why won't it export". Refuses, with the reason named, on a missing verdict, a FAIL verdict, a changed list, or a verdict that has gone stale. (richapi-gtm)

  Self-containedDeclared tools2,825 words
• learn

  The local learnings flywheel. Records aggregate provider and hop hit rates from a run journal into gtm/learnings.jsonl, and applies them as decayed, attributed priors so the next run starts better than the last. Use when asked "remember what worked", "which provider wins", "apply what we learned", "reorder the waterfall", or "what do we know so far". Strictly local: no server, no upload, no consent toggle, no network call of any kind. Makes zero API calls. (richapi-gtm)

  Self-containedDeclared tools2,575 words
• list-hygiene

  Dedupe, validate and suppress a contact list before anyone spends enrichment credits on it — duplicate collapse, normalisation, email classification, domain liveness, and a fail-closed do-not-contact cross-check. Use when asked to "clean this list", "dedupe this", "prep it for send", "verify before campaign", "check for dead domains", or "is this list any good". Proactively suggest before /enrich-waterfall: enriching a dirty list pays full price for rows that were never going to work. (richapi-gtm)

  Self-containedDeclared tools3,214 words
• local-business-prospecting

  Sources brick-and-mortar and SMB prospects from maps and directories rather than from LinkedIn — "what and where" instead of "title and industry". Google Maps by keyword or by a Maps URL, Yellow Pages for the long tail, and review mining for the outreach hook. Every one of its endpoints is billed per result and none of them reports its charge, so the plan is the only place the bill is ever visible. Use when asked "find all dentists in Austin", "list of restaurants near", "local business list", "scrape Google Maps for", "businesses in <city> with <criterion>", or "SMBs with no website". (richapi-gtm)

  Self-containedDeclared tools5,197 words
• measure

  Reports what a GTM run actually did — coverage first, then spend as an honest range, then hit rates by hop and provider. Use when asked "how did that run go", "what did that cost", "what's my hit rate", "credits per meeting", "measure this play", or "was that worth it". Reads the ledger and the run journal; makes zero API calls and spends nothing. Can also write an aggregate-only artifact that is safe to forward. (richapi-gtm)

  Self-containedDeclared tools3,529 words
• org-map

  Maps the buying committee at one account (who reports to whom, who influences, who signs), as a graph where every edge carries its provenance and a line nobody observed is drawn as a line nobody observed. Use when asked to "map the org", "who reports to whom at X", "build the buying committee", "who is the economic buyer", "who signs this", "find the champion", or "who is the new SDR's manager". Proactively suggest after /account-research surfaces a shortlist and before multithreading an account. Runs local inference by default; calls ai_enrich only for web grounding or batch scale. (richapi-gtm)

  Self-containedDeclared tools3,900 words
• outreach-expert

  Advises on outreach setup up to (and only up to) the send button: domain and mailbox strategy, SPF/DKIM/DMARC, warmup, sender rotation, deliverability diagnosis, bounce handling and sequence hygiene. Use when asked "my emails go to spam", "how do I warm up a domain", "SPF DKIM DMARC", "how many mailboxes", "sender rotation", "how many steps in the sequence", "bounces are climbing", or "review my sending setup". It never sends, warms, or configures anything — sending execution is outside this pack permanently. (richapi-gtm)

  Self-containedDeclared tools3,399 words
• personalize

  Writes personalised outreach (first lines and short email bodies) grounded in a research brief, where every claim traces to a source line and an unsupported claim is refused rather than written. Use when asked to "personalise this", "write a first line", "draft an opener", "write cold emails for this list", "make this sound less generic", or "reference their funding round". Proactively invoke after /evidence-score and before /sequence-builder. Runs local inference by default; calls ai_enrich only for web grounding or batch scale. (richapi-gtm)

  Self-containedDeclared tools2,580 words
• play-design

  Designs a repeatable GTM play — the trigger that starts it, the audience it applies to, the sequence of existing skills that runs, and the number that says whether it worked. Writes a named play record the team can save, re-run and measure; composes /signal-watch, /enrich-waterfall, /campaign-review and /measure rather than reimplementing any of them. Use when asked to "design a play", "turn this into a repeatable motion", "we should do this every time X happens", "productise this campaign", "build us a playbook", or "how do we run this again next month". (richapi-gtm)

  Self-containedDeclared tools2,905 words
• pre-meeting-briefing

  A call sheet on the humans you are about to meet, built inside a ten-minute box and a plan you approve once. Person-first, not company-first: who is in the room, what changed at their company since anyone last looked, and one honest opener. Use when asked to "prep me for my meeting", "brief me on <name>", "call with X tomorrow", "who am I meeting", "who is this person", or "interview prep". It owns no endpoints and fetches nothing itself — it constrains and composes /account-research and /enrich-waterfall, and it never runs the market pass. (richapi-gtm)

  Self-containedDeclared tools3,088 words
• reply-triage

  Classifies inbound replies pasted or exported out of a sender (interested, objection, not now, wrong person, unsubscribe) and routes each one. The opt-out path is a compliance gate rather than a category: it runs before classification, it is decided locally, it fails closed, and it writes through the pack's real suppression store, so an ambiguous reply is suppressed instead of contacted again. Classification itself is free local inference; the paid LLM hop is reserved for batch scale and never touches the opt-out decision. Use when asked to "triage these replies", "what did they say", "process my inbox", "handle unsubscribes", "who is interested", or "sort these responses". (richapi-gtm)

  Self-containedDeclared tools3,011 words
• research-agent

  Answers a freeform research question across a list — the escape hatch for questions no specific skill covers. Recognises the question shape first, routes it to a known endpoint path with a known cost, shows the fan-out per row AND the list total before anything runs, and returns the explicit null for a question whose answer is not findable. Use when asked "find out X for each of these companies", "research this question across my list", "can you check whether they...", "I need a custom column", or when a custom research-column prompt from another enrichment tool is pasted. Runs local inference; calls ai_enrich only for Perplexity web grounding. (richapi-gtm)

  Self-containedDeclared tools5,005 words
• richapi-gtm

  Entry point for the RichAPI GTM pack. Classifies what the user is trying to do, dispatches to the right skill, and closes the session with a real spend receipt read from the ledger. Use when asked "what can you do with RichAPI", "help me with outbound", "I have a list of leads", or any GTM request that does not obviously match one skill. Proactively invoke instead of answering a GTM data question directly. (richapi-gtm)

  Self-containedDeclared tools2,605 words
• scheduled-workflow

  Runs a saved workflow on a schedule, under an envelope the user approved in advance for a bounded, named set of work. Use when asked to "run this every week", "schedule this", "automate this refresh", "do this nightly", or "set up a recurring enrichment". The scheduled run re-derives the plan before it spends, compares it against what was approved, and STOPS on any divergence — a repricing, a bigger plan, an exhausted envelope, an expired approval, or any gate that would have asked a human. Makes zero API calls itself. (richapi-gtm)

  Self-containedDeclared tools6,560 words
• sequence-builder

  Designs the outreach sequence (how many steps, how far apart, on which channel, and what each step is for) and writes a copy skeleton with the merge-tag syntax of the sender the team already uses. Spends nothing: sequence design is thinking, not fetching. Use when asked to "design a sequence", "build a cadence", "how many follow-ups", "what should step three say", "space these touches out", or "turn these drafts into a campaign". Hands off to /launch, which is the only skill that produces the artifact a sending tool ingests. (richapi-gtm)

  Self-containedDeclared tools2,722 words
• signal-watch

  Watches a list of accounts for the moment they become buyable (hiring, ads, posts, tech-stack moves, funding, news, champion departures) and writes each change to `triggers.jsonl` as a dated digest of diffs rather than a wall of static facts. The only recurring-cost skill in the pack: it prices the whole standing charge per cycle, per week and per month before the first cycle runs, and re-prices it when the catalog moves. Also runs in customer mode over accounts that already pay you, for expansion and churn. Use when asked to "monitor these accounts", "watch this list", "buying signals", "hiring intent", "trigger digest", "alert me when X starts hiring", "who is about to churn", or "what changed this week". (richapi-gtm)

  Self-containedDeclared tools5,447 words
• tam-map

  Sizes a total addressable market and breaks it into segments the rest of the pipeline can act on — reading the count from the one endpoint that reports a total for free rather than paging until the answer appears on the invoice. Use when asked "how big is this market", "how many companies match our ICP", "size the TAM", "segment the market", "how many SMBs in this city", "build the account universe", or "find lookalikes for these customers". Proactively suggest after /icp-review and before /build-prospect-list: sizing the universe is cheaper than enumerating it. (richapi-gtm)

  Self-containedDeclared tools4,229 words