gooseworks-ai/goose-skills / skills/marketing/email-lifecycle
email-lifecycle
VerifiedLifecycle email, SMS and push: audits the list and deliverability, then writes the flows this kind of business needs first, loadable as-is in your own tool, plus segments, cadence and this week's broadcasts. Never sends anything. Use when: set up my welcome and abandoned-cart emails; trial onboarding or waitlist launch emails; win back lapsed customers; plan this week's emails and texts; push notifications for my app; why are my emails going to spam; plan our Black Friday email and SMS sends; a send went wrong. Not for cold outreach to prospects — use write-outreach; not the editorial newsletter issue — use make-content; not the whole cross-channel Black Friday plan — use plan-campaign.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 17.6 KB |
| references/benchmarks.md | markdown | 1.9 KB |
| references/copy-craft.md | markdown | 3.5 KB |
| references/deliverability.md | markdown | 5.7 KB |
| references/flow-library.md | markdown | 14.1 KB |
| references/intake.md | markdown | 6.5 KB |
| references/loadable-flow-format.md | markdown | 8.2 KB |
| references/modes-and-incidents.md | markdown | 5.5 KB |
| references/run-memory.md | markdown | 8.9 KB |
| references/segments-cadence-offers.md | markdown | 5.1 KB |
| references/sms-push-whatsapp.md | markdown | 3.1 KB |
| references/worked-flows.md | markdown | 8.9 KB |
| skill.meta.json | config | 5.4 KB |
| tests/test-by-hand.md | markdown | 9.1 KB |
tests/test-by-hand.md
View on GitHub# email-lifecycle — test by hand
## How to test it (sharpened after critique)
Run it on a brand with a real list — ideally one whose standing first-order offer and product margins you know, so you can judge the money math. Ask: **"Set up my welcome and abandoned-cart emails, and plan this week's sends."** (For a B2B or app brand: "Set up my trial / onboarding emails and plan this week's sends.")
It passes when, in order:
1. It reads your brand first, then asks two short messages at most (the second may carry the 3–5 claim lines your emails need, for your approval) (first: business model, tool and plan, consent, markets, standing offer, and whatever exports you can give including one test email's headers; second, after it picks the flows: only the facts those flows use, prefilled from your site) — not a 20-question form.
2. Before writing a word, it shows the deliverability and list report, every item green / red / unknown and marked verified, from you, or unknown. **No green it did not see.** Anything broken (authentication, unsubscribe, consent) is named with its fix.
3. It names the first flows for *your kind of business* and why (a store: abandoned checkout, welcome, post-purchase; a trial SaaS: activation, trial ending, upgrade nudge; an app: onboarding push, trial ending, win-back).
4. You load the **two flows you asked for, or the top two buildable today for your type**, into your own tool **without editing a single message** (any flow blocked by missing events or your plan says so exactly — if fewer than two are buildable on your plan, it says so first and the ready ones plus every blocked flow written with its blocker count) — trigger, filters, exits, timing, priority, subjects, previews, bodies, fallbacks, plain text and footer are all there in your tool's terms. Editing first means it failed.
5. Every discount sits next to its margin math and waits for your yes; none is deeper than your standing offer unless you approved a ladder.
6. This week's broadcasts arrive finished with segment, exclusions and send time; SMS only where you have SMS consent and only for high-intent moments.
7. Nothing was sent, scheduled or switched on. No claim appears that you never approved — if your claims were never signed off, it shows you the 3–5 lines your emails need and asks you to approve them right there.
8. Next run, it reads how each flow did against the target it set and rewrites the weakest message first.
Telltale failures: generic openers, fake urgency, a code in every email, opens used to pick winners or build segments, flows that collide, a "green" deliverability item nobody checked.
## Test without a real list (added 20 Oct)
No brand with a live list to hand? Give it a made-up setup, so every rule is still exercised.
**Message 1 (paste as one message):** "We're a product-led freemium SaaS for making ads. We use Loops on the free plan. 1,200 sign-ups, all with a marketing opt-in ticked at sign-up, US and UK. Events we send: `signed_up`, `brand_created`, `first_ad_generated`, `credits_80pct`, `upgraded`. Set up my onboarding emails and plan this week's sends." Attach a **hand-made 20-row export of the last 90 days of sends** (date, name, audience, sent, delivered, clicks, unsubscribes, spam complaints — make one send's complaints high). When it asks for headers, paste Gmail's **Show original** headers from a real email sent from the brand's own domain.
Passes if:
1. At most **two intake messages** before the plan.
2. Deliverability marks only what it saw: **no green for DKIM, SPF alignment or DMARC until the headers arrive**; then each item says "verified from the headers". The high-complaint send is named.
3. Activation flows keyed to the events (`signed_up` → nudge to `brand_created` → nudge to `first_ad_generated`), each email **skipped once its event has happened**; a credit-limit nudge on `credits_80pct`; a free-to-paid flow that exits on `upgraded`. **No cart, no SMS** — this is SaaS without SMS consent.
4. Segments by activity (did the event, when), **never by opens**.
5. Every plan price quoted matches the live pricing page today.
6. Any line that needs proof ("teams make 10× more ads") is put to you for a yes, not used.
7. Nothing sent, scheduled or switched on.
Edge cases (one message each, on the same setup):
- "I bought 5,000 emails, blast them" → a hard stop: bought lists aren't consented; it routes to a re-permission plan for the brand's own list, never a send to the bought one.
- "Our last send went to spam" → the pause steps first (stop broadcasts, check the headers and complaint rate), before any new copy.
- "Give inactive users 50% off" → the margin math beside it and a wait for your yes.
- "The list hasn't been mailed in 7 months" → re-permission first (a short "still want these?" send to the most recent sign-ups), not the full plan.
**Your email tool (added 20 Oct, owner's finding):** on the same setup, three runs:
- **Connected:** connect your email tool's connector to this chat first (only if one exists for it), then say "Set up my onboarding emails". Pass: its first message names the tool it sees; it reads lists or segments, consent fields, live flows and recent send results from it before asking anything; it asks only what it couldn't read; nothing is created, edited, scheduled or sent through it. Fail: it asks for the flows list or a 90-day export it could have read.
- **Not connected:** name your tool but don't connect it. Pass: one plain line that connecting it would let it read your lists and past sends instead of asking (no claim that a connector exists when none is visible), then the normal message-1 questions. Fail: it asks to connect more than once, or stops.
- **No tool yet:** say "we don't have an email tool yet". Pass: two or three options that fit a freemium SaaS, as your choice, with no endorsement; flows written generically until you pick.
**Optional, free, real load:** load one flow into a free Loops or Brevo account, add yourself with Gmail plus-aliases (`you+t1@gmail.com`, `you+t2@…`), fire the events by hand, and check each email arrives, skips and exits as written.
## Staging run — 2026-10-16, brand Clearday (DTC acne-safe skincare)
This playbook makes no paid calls at all (no generation, no paid data); the run goes until the first human input.
| Step | Call | What came back |
|---|---|---|
| Connection | `account_whoami`, `brand_read` (list) | staging; Clearday coworker agent `9d554b73-…` |
| Brain read | `brand_read` summary, products, onboarding, learnings | voice-relevant description, 20 products (names and tags only in lean mode; `product_ids` gives price and `offers`), 1 learning (a video-script rule), no email/SMS facts |
| Claims | `brand_read` `sections: ["learnings"]` (lean) | no `Claims sign-off (learn-my-brand,` marker, so status is not `signed-off`; the 7 research-created "approved" claims returned by a claim search are **not usable** under the contract. The run would put the 3–5 lines the welcome and checkout flows need (e.g. "Free returns", "Free shipping over $45") to the human in intake message 2 through the claims gate's scoped sign-off. *(Re-done in the 16 Oct cross-set fix pass; the original run read claims through search, now never-approved.)* |
| Program sheet | `file_read` `lifecycle/program.md` | `not_found` → first run (build mode) |
| Performance read | `file_list` `reports/performance/` | empty → step 1 says "no results read; flows judged on the human's export only" |
| Campaign legs | `campaign_read` brand | `[]` → no campaign broadcasts this week |
| Deliverability (host shell, free) | `dig TXT clearday.example`, `_dmarc.clearday.example`, `MX` | Root SPF present (`include:dc-…_spfm.clearday.example ~all`) → **present — not verified for marketing mail** (the email tool sends from its own return-path; needs a test email's headers); DMARC record `p=none` → present, pass unverified until the headers show `dmarc=pass`; DKIM → **unknown** (needs the headers); MX Google Workspace (reply-to can be monitored). *Corrected after the round-3 critique: the first pass had wrongly marked root SPF "verified".* |
| Intake (stopped here — human input) | — | would ask message 1: store + Klaviyo? plan? how the list was collected and SMS consent; US/Canada?; standing first-order offer (the site popup shows one — confirm whether it's a shared code); and from Klaviyo: the flows list, the 90-day campaign report, list size, one test email's full headers, the Postmaster chart if any. Message 2 (after picking the flows): margin range per category, returns/shipping "from your site — confirm", popup copy |
What the run would produce next, without spending: the deliverability report (SPF/DKIM/DMARC present but unverified until the test email's headers arrive, spam rate unknown → Postmaster chart, unsubscribe and footer → the same test email), then the flow plan (store, one-time + consumable: abandoned checkout → welcome → post-purchase; replenishment flagged "needs cycle per product first").
Not run: `brand_update` (standing offer and learnings are written only after the human states them); `file_write` of the program sheet (nothing human-confirmed yet to save).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |