Eight complete programs, each solving one recognisable business problem, each runnable exactly as printed (the outputs below are their real runs). Adapt by editing the quoted strings — names, messages, thresholds — and re-running.
1 · The classic: score-and-route triage
The poster program, grown up: three-way routing with a guard for unreachable leads.
when a new lead arrives score it with ai if phone is "" then create task "Find contact info for {name}" and stop if score is hot then assign to "Maya" and reply on whatsapp "Thanks {name}! We'll call you in 10 minutes." otherwise if score is warm then assign to "Sam" and add tag "follow-up-today" otherwise add tag "nurture"
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation ● event: lead.created → name: Ada · phone: +1 555 0100 · email: [email protected] ⚡ score it with ai → score 84 · HOT ✗ if phone is "" → no ✓ if score is hot → yes → assign to "Maya" → lead assigned ✉ reply on whatsapp → "Thanks Ada! We'll call you in 10 minutes." ✔ done in 2.1ms — 1 message · assigned to Maya
2 · Form follow-up with a task trail
when a form is submitted score it with ai add tag "form: {form}" if score is hot then assign to "Maya" send email to "[email protected]" "Hot form lead: {name} — {message}" create task "Reply personally within the hour" for "Maya" otherwise reply on email "Thanks {name} — we'll be in touch this week."
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation ● event: form.submitted → form: Contact us · name: Grace Hopper · email: [email protected] ⚡ score it with ai → score 88 · HOT ⚑ add tag "form: Contact us" → tagged ✓ if score is hot → yes → assign to "Maya" → submission assigned ✉ send email to "[email protected]" → "Hot form lead: Grace Hopper — Interested in the Pro plan" ☑ create task "Reply personally within the hour" → created ✔ done in 2.1ms — 1 message · assigned to Maya · tags: form: Contact us · 1 task
3 · A drip sequence
wait makes time a step. Locally the waits are instant and labelled; in production this is a real three-touch sequence.
when a new lead arrives reply on email "Welcome, {name}! Here's what Skode can do." wait 2 days reply on email "A quick case study you might like, {name}." wait 3 days reply on email "Any questions? Just hit reply." add tag "drip-complete"
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation ● event: lead.created → name: Priya · phone: +1 555 0100 · email: [email protected] ✉ reply on email → "Welcome, Priya! Here's what Skode can do." ◌ wait 2 days → (simulated — instant here, real delay in Skode) ✉ reply on email → "A quick case study you might like, Priya." ◌ wait 3 days → (simulated — instant here, real delay in Skode) ✉ reply on email → "Any questions? Just hit reply." ⚑ add tag "drip-complete" → tagged ✔ done in 2.1ms — 3 messages · tags: drip-complete
4 · Invoice receipts, with a VIP touch
when an invoice is paid reply on email "Receipt for invoice {number} — thank you!" add tag "paid" if amount is above 5000 then create task "Send a handwritten thank-you to {name}" for "Maya" add tag "vip"
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation
● event: invoice.paid → number: INV-0461 · amount: 9500 · name: Margaret H
✉ reply on email → "Receipt for invoice INV-0461 — thank you!"
⚑ add tag "paid" → tagged
✓ if amount above 5000 → yes
☑ create task "Send a handwritten thank-you to {name}" → created
⚑ add tag "vip" → tagged
✔ done in 2.1ms — 1 message · tags: paid, vip · 1 task5 · Message keyword router
Sequential ifs make a tidy keyword switchboard; functions keep each route readable.
when a message arrives if text contains "pricing" then route to sales if text contains "refund" then route to support if text contains "partner" then create task "Partnership inquiry from {name}" for "Ops" to route to sales assign to "Maya" and add tag "sales-inquiry" reply on whatsapp "Great question, {name} — connecting you with Maya now." to route to support assign to "Sam" and add tag "support" reply on whatsapp "Sorry to hear that, {name} — Sam is on it."
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation ● event: message.received → from: +1 555 0101 · name: Lin · channel: whatsapp ✓ if text contains "pricing" → yes ƒ route to sales → run your function → assign to "Maya" → message assigned ⚑ add tag "sales-inquiry" → tagged ✉ reply on whatsapp → "Great question, Lin — connecting you with Maya now." ✗ if text contains "refund" → no ✗ if text contains "partner" → no ✔ done in 2.1ms — 1 message · assigned to Maya · tags: sales-inquiry
6 · Deal-won celebration and handover
when a deal is won send slack "{owner} closed {name} for {amount}!" create task "Kick off onboarding for {name}" for "Success Team" add tag "customer" set commission to amount / 20 log "commission earned: {commission}"
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation
● event: deal.won → name: Skylark Corp · amount: 14000 · owner: Maya
✉ send slack → "Maya closed Skylark Corp for 14000!"
☑ create task "Kick off onboarding for {name}" → created
⚑ add tag "customer" → tagged
✎ set commission → = 700
• log → commission earned: 700
✔ done in 2.1ms — 1 message · tags: customer · 1 task7 · Multi-channel escalation sweep
The pattern for "make sure someone sees this": alert on every channel, then log an audit line. Note that a channel is part of the sentence, not a value — send slack … — so each channel gets its own explicit send (a feature, for auditability: the reader sees exactly where alerts go). A function keeps the sweep tidy, and a small loop assembles the audit trail:
when a new lead arrives score it with ai if score is above 95 then sound the alarm create task "Drop everything and call {name}" for "Maya" set alerted to "" for each ch in "slack", "email", "sms" add "{ch} " to alerted log "alerted via: {alerted}" to sound the alarm send slack "URGENT: {name}, budget {budget} — score {score}" send email to "[email protected]" "URGENT: {name}, budget {budget}" send sms "URGENT lead: {name}"
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation ● event: lead.created → name: Whale Corp · phone: +1 555 0100 · email: [email protected] ⚡ score it with ai → score 97 · HOT ✓ if score above 95 → yes ƒ sound the alarm → run your function ✉ send slack → "URGENT: Whale Corp, budget 80000 — score 97" ✉ send email to "[email protected]" → "URGENT: Whale Corp, budget 80000" ✉ send sms → "URGENT lead: Whale Corp" ☑ create task "Drop everything and call {name}" → created ✎ set alerted → = "" ↻ for each ch in ["slack", "email", "sms"] → 3 items ✎ add to alerted → = "slack " ✎ add to alerted → = "slack email " ✎ add to alerted → = "slack email sms " • log → alerted via: slack email sms ✔ done in 2.1ms — 3 messages · 1 task
8 · The self-auditing rule
The pattern that makes on-call life bearable: narrate decisions as you make them, and end with a summary a human can read in the run log.
when a new lead arrives score it with ai set verdict to "undecided" if score is hot then set verdict to "routed to Maya" and assign to "Maya" if score is warm then if budget is above 500 then set verdict to "warm but funded - Sam" and assign to "Sam" otherwise set verdict to "nurture list" and add tag "nurture" if score is cold then set verdict to "archived" and add tag "cold" log "AUDIT {name} from {source}: score {score} ({heat}) -> {verdict}"
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation ● event: lead.created → name: Kiran · phone: +1 555 0100 · email: [email protected] ⚡ score it with ai → score 55 · WARM ✎ set verdict → = "undecided" ✗ if score is hot → no ✓ if score is warm → yes ✓ if budget above 500 → yes ✎ set verdict → = "warm but funded - Sam" → assign to "Sam" → lead assigned ✗ if score is cold → no • log → AUDIT Kiran from webinar: score 55 (warm) -> warm but funded - Sam ✔ done in 2.1ms — assigned to Sam
9 · Budget-band pricing assistant
Thresholds as data: one list, one loop, one decision — change the bands without touching the logic.
when a new lead arrives set band to "starter" for each threshold in 1000, 2500, 10000 if budget is at least threshold then set band to "band-{threshold}" add tag "{band}" log "{name} with budget {budget} lands in {band}"
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation ● event: lead.created → name: Kiran · phone: +1 555 0100 · email: [email protected] ✎ set band → = "starter" ↻ for each threshold in [1000, 2500, 10000] → 3 items ✗ if budget is "at least threshold" → no ✗ if budget is "at least threshold" → no ✗ if budget is "at least threshold" → no ⚑ add tag "starter" → tagged • log → Kiran with budget 2600 lands in starter ✔ done in 2.1ms — tags: starter
(A subtlety worth stealing: the loop runs low-to-high and simply overwrites band at each threshold the budget clears, so the final value is the highest band cleared — no nested ifs required. Note add tag "{band}" interpolating a variable into the label.)
10 · Weekly-source hygiene check
Not every skript talks to customers. This one polices data quality and files the paperwork for a human — automation as janitor.
when a new lead arrives set problems to "" if source is "" then add "no-source " to problems if email is "" then add "no-email " to problems if problems is not "" then add tag "needs-cleanup" create task "Fix lead data for {name}: {problems}" for "Ops" log "flagged: {problems}" otherwise log "clean record"
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation
● event: lead.created → name: Ghost Lead · phone: +1 555 0100 · email:
✎ set problems → = ""
✓ if source is "" → yes
✎ add to problems → = "no-source "
✓ if email is "" → yes
✎ add to problems → = "no-source no-email "
✓ if problems is not "" → yes
⚑ add tag "needs-cleanup" → tagged
☑ create task "Fix lead data for {name}: {problems}" → created
• log → flagged: no-source no-email
✔ done in 2.1ms — tags: needs-cleanup · 1 task11 · The polite decliner
Saying no is an automation too. Small budgets get a kind message and a future-proof tag instead of silence:
when a new lead arrives score it with ai if budget is below 100 then reply on email "Thanks for your interest, {name}!" "We're likely overkill for your current needs - but keep us in mind as you grow." add tag "future-fit" and stop add tag "qualified"
output — real, executed at build time
▶ example.skript — Skode Skript · local simulation ● event: lead.created → name: Tiny Co · phone: +1 555 0100 · email: [email protected] ⚡ score it with ai → score 30 · COLD ✓ if budget below 100 → yes ✉ reply on email → "Thanks for your interest, Tiny Co!\nWe're likely overkill for your current needs - but keep us in mind as you grow." ⚑ add tag "future-fit" → tagged ■ stop → skript halted ✔ done in 2.1ms — 1 message · tags: future-fit
Every recipe here is four ideas deep at most — an event, a decision, a handful of actions, and a note for the humans. That is not because the examples are simplified; it is because that is the honest shape of good automation. When yours grows past it, reach for a function and give the complexity a name.