Lead Nurture Sequencer
Nurture cadences break in one of two ways: leads go quiet because nothing re-engages them, or automation keeps emailing someone who already replied, booked a meeting, or unsubscribed. This skill fixes both -- it stages every lead off real signal, not off "days since the sequence started," and hard-stops automation the moment a human should own the thread.
When to use
- Deciding which leads are due for a nurture touch today, and what that touch should be.
- Auditing an existing sequence for suppression bugs -- leads who replied or unsubscribed but are still receiving automated sends.
- Building the staging logic behind an email/marketing-automation nurture flow.
When NOT to use
- For lead scoring against an ICP (fit, firmographics) -- that's a qualification skill, not a nurture-cadence one. This only reads engagement behaviour.
- For writing the actual nurture email copy -- this decides what kind of touch and when, not the words.
Method
- Check suppression first, always:
unsubscribed and meeting_booked short-circuit everything else -- no engagement math overrides them.
- A real
replied beats every automated signal -- a human must respond within 24 hours, and no scheduled send should follow.
- Absent those, read opens/clicks/recency: clicks + recent activity = warm (send proof); opens with no clicks = curious (vary the angle); no opens for 30+ days = cold (breakup email, don't just keep silently sending).
- Attach a cooldown per stage and compute how many days until the next touch is due, so the plan is schedulable, not just descriptive.
Inputs
brand -- the brand nameleads[] -- each {name, days_since_last_touch, email_opens, email_clicks, replied, meeting_booked, unsubscribed}. In production these come from the email platform or CRM webhook; the demo ships a deterministic sample.
Output (JSON)
leads[] (each with name, stage, action, next_touch_in_days, due_now), stage_counts, due_now[] (names due today), and a one-line summary.
Run it
python scripts/nurture_sequencer.py # built-in sample
python scripts/nurture_sequencer.py in.json # your own brand + leads
Zero dependencies, no API keys. To go live, swap the sampled leads for a real CRM/email-platform pull; the staging and suppression logic stays identical. A scheduled n8n version of this exact logic follows the same pattern as pipeline-automation/aeo-geo-improvement-loop.json.
1---2name: lead-nurture-sequencer3description: Stages leads by real engagement signal (opens, clicks, replies, meetings booked, unsubscribes, days since last touch) and decides the next nurture action for each -- what to send, and how many days until it's due. Enforces the suppression/cooling rules that keep automated nurture from talking over a real reply or a booked meeting, or re-touching someone who unsubscribed. Use to run or audit an email/lifecycle nurture cadence, decide who's due for a touch today, or check that automation isn't stepping on human-handled leads.4---56# Lead Nurture Sequencer78Nurture cadences break in one of two ways: leads go quiet because nothing re-engages them, or automation keeps emailing someone who already replied, booked a meeting, or unsubscribed. This skill fixes both -- it stages every lead off real signal, not off "days since the sequence started," and hard-stops automation the moment a human should own the thread.910## When to use1112- Deciding which leads are due for a nurture touch today, and what that touch should be.13- Auditing an existing sequence for suppression bugs -- leads who replied or unsubscribed but are still receiving automated sends.14- Building the staging logic behind an email/marketing-automation nurture flow.1516## When NOT to use1718- For lead scoring against an ICP (fit, firmographics) -- that's a qualification skill, not a nurture-cadence one. This only reads engagement behaviour.19- For writing the actual nurture email copy -- this decides *what kind* of touch and *when*, not the words.2021## Method22231. Check suppression first, always: `unsubscribed` and `meeting_booked` short-circuit everything else -- no engagement math overrides them.242. A real `replied` beats every automated signal -- a human must respond within 24 hours, and no scheduled send should follow.253. Absent those, read opens/clicks/recency: clicks + recent activity = warm (send proof); opens with no clicks = curious (vary the angle); no opens for 30+ days = cold (breakup email, don't just keep silently sending).264. Attach a cooldown per stage and compute how many days until the next touch is due, so the plan is schedulable, not just descriptive.2728## Inputs2930- `brand` -- the brand name31- `leads[]` -- each `{name, days_since_last_touch, email_opens, email_clicks, replied, meeting_booked, unsubscribed}`. In production these come from the email platform or CRM webhook; the demo ships a deterministic sample.3233## Output (JSON)3435`leads[]` (each with `name`, `stage`, `action`, `next_touch_in_days`, `due_now`), `stage_counts`, `due_now[]` (names due today), and a one-line `summary`.3637## Run it3839```bash40python scripts/nurture_sequencer.py # built-in sample41python scripts/nurture_sequencer.py in.json # your own brand + leads42```4344Zero dependencies, no API keys. To go live, swap the sampled `leads` for a real CRM/email-platform pull; the staging and suppression logic stays identical. A scheduled n8n version of this exact logic follows the same pattern as `pipeline-automation/aeo-geo-improvement-loop.json`.
saurabhshuklagrowisto