Setup & Customization Checklist

1. Trigger setup

This skill is designed to fire on account signal events — website visits, registrations, trial starts, usage milestones, meeting completions, outreach replies, LinkedIn engagement — and to be callable on demand ("score this account"). Reference trigger instructions:

An account signal fired: [event type] for [company / domain].

Load and follow the <Account tier scoring> skill in full.

Context for this run:
- Triggering event: [type + payload summary]
- Named individual (if any): [name, title]

• Wire your signal sources (web analytics/identification, product events, CRM activity, LinkedIn engagement feed) to invoke the skill with the event context.
• Pass the event type through — several rules branch on it (the partner-check exception, the self-artifact guard, agency flagging all key off signup/trial events).
• Account memory must persist between runs — the methodology scores the running signal stack, not single events. Without durable memory, stacking rules and repeat-view escalation cannot work.

2. Placeholders

• {{CRM}}, {{ALERTS_CHANNEL}}, {{ACTIVE_DEALS_CHANNEL}}, {{SALES_OWNER}}, {{PARTNERSHIPS_OWNER}}, {{SOCIAL_SELLER}}, {{SELF_SERVE_THRESHOLD}}, {{ENTERPRISE_THRESHOLD}}, {{BUYER_PERSONA_LIST}}, {{DISQUALIFIED_MARKETS}} — all filled (or the optional ones deliberately deleted).
• Map the lifecycle references to your funnel stage names: closed won, trial, closed lost, self-serve, unqualified, potential partner. The skill references stage semantics, not stage names.
• Create the tags in your workspace: the four tier tags, the three ACV tags, self-serve, expansion target, and confirm your vendor/partner labels match what Step 0a checks.
• Create the lead-score field in {{CRM}} that Step 4 syncs to, with the four tier values.
• Tier names: Bronze/Silver/Gold/Diamond is the default ladder — rename freely, but keep four levels or re-map every rule that names a tier.
• If you have no LinkedIn signal feed, delete the {{SOCIAL_SELLER}} signal rows and the LinkedIn routing caps — do not leave them referencing a feed that doesn't exist.

3. Policy decisions

• Start in EVAL mode. Run 1–2 weeks with CRM writes disabled and review the would-have-written tags in the alert posts before going live.
• All numeric thresholds are tuned defaults, not laws: the self-serve floor (<50 employees AND ≤$10M funding), the enterprise gate (1,000+), the <3-sellers cap and its overrides (>$5M revenue founder signup; >$10M funding), the fresh-trial window (14 days), the inactivity window (30 days), the funding bands in the ACV assessment. Recalibrate each to your price point and sales-touch economics — then leave them stable so tiers stay comparable over time.
• The Silver→Gold tiebreaker (default to visibility on genuinely mixed signals) — confirm {{SALES_OWNER}} actually wants the extra review volume; flip the default if your alert channel is drowning.
• The prior-relationship Gold floor — decide what counts as a "genuine prior evaluation" in your CRM (deal object, logged meeting, trial record) and confirm the verification signals exist in your data.
• The ACV assessment dimensions are written for a GTM/sales product ("dedicated sellers", "RevOps hires"). Substitute the team and ops function your product serves in dimensions 2, 3, and 6.
• Alert format: Gold/Diamond posts use your team's standard lead-scoring alert format — point Step 7 at that spec; never let this skill invent a second layout for the same channel.

4. Behavioral invariants (do not remove)

• The self-serve gate runs before everything and is a hard stop — no signal strength overrides headcount.
• Near the self-serve threshold, headcount is verified against LinkedIn, taking the lower of the two counts — enrichment ranges are never trusted alone in the risk band.
• Exactly one tier tag and one ACV tag per account — old tags removed on every write.
• Every tag write is verified by re-reading the record, retried once, and reported honestly — an alert never claims a write that didn't land (TIER_TAG_STATUS).
• Bronze and Silver never alert. Zero-signal accounts never alert.
• Anonymous signals never score Gold; post-engagement-only never scores Gold; a lone profile view caps at Silver.
• A weak new signal never demotes an existing tier — demotion only via explicit gates or deliberate review.
• The self-artifact guard: objects created by the current run never count as evidence a rep is already engaged.
• Auto-created onboarding objects never count as product depth.
• All outreach produced by this skill is queued for human approval — never auto-sent, at any tier.
• Memory snapshots prepend, never overwrite — the running signal stack is the scoring substrate.