Research report · 11 October 2026 · Run 21

A welcome-email brief is a draft, not a configured flow

Keep Core Haines emails as a conditional source-inspected candidate for our proposed five-email post-signup welcome draft. Start with an owner-maintained brief and reviewed prompt. Neither route was run or compared.

What was inspected

On 11 October, the mutable selected emails/SKILL.md displayed 2.1.3. The returned selected text was 313 lines and marked crawled today. I also read its directly routed welcome sequence template and per-email copy guidelines, returned as crawled four and two days earlier. These mutable pages are bounded observations, not an immutable selected-file pin or a complete package inventory.

No path-history/API retry, directory enumeration, installation, local product-context read, email account connection or send occurred. The selected-file SHA/date, latest modifying commit, full 2.1.1→2.1.3 diff and current reference inventory remain unknown. The file-attributable 2.1.0 change commit remains historical; it is not a latest main pin.

Separate source fields from pool controls

Field	Inspected source	Public decision ceiling
Job and length	Post-signup welcome is generally 3–7 emails; the source illustrates 5–7 over 12–14 days and the reference has seven sample messages. Product-user onboarding is a separate job.	Five emails is our subset for a post-signup welcome draft, not a source requirement or validated cadence.
Inputs	Asks for audience, trigger, relationship, goals, other emails and performance; instructs reading local product-marketing context if present.	Use only owner-approved, non-personal context. Mark other emails and performance unknown if not supplied. Do not read local context or customer records without approval.
Timing and exits	Supplies example timing and asks for timing and exit-condition fields; per-email segment/conditions is “if applicable.”	Record proposed delays, overlap and exit questions as hypotheses. The recipient-policy and ESP owners determine eligibility, consent, suppression, frequency, timezone and actual rules separately.
Copy and CTA	Template suggests result, objection, offer and urgency beats. Source asks for body, a main CTA and destination.	Omit or hold unsupported result, case study, urgency, guarantee or promise. A named owner supplies approved proof and terms and checks the exact CTA URL before handoff.
Metrics	Requests measures and “benchmarks”; the copy reference includes generic rate ranges without cited audience, denominator or method.	Metric ideas only. The owner supplies definitions and an authorized aggregate baseline, or records “not supplied.” Do not publish generic rates as targets or claim lift, conversion or deliverability.
Implementation	Routes to the two opened references, an unreviewed Copywriting AI-tells reference and a separate mutable tools registry. A Markdown draft requires no selected-text ESP action.	Cross-skill and tool routes remain unreviewed and unauthorized. A draft grants no ESP, test or send authority.

The source requests an initial assessment, a sequence overview, per-email content and a metrics plan. It does not enforce our five-email limit, proof IDs, hold markers, named sign-off, baseline definition or separate ESP release. The source's prescriptive timing and rate language has no cited campaign comparison or recipient-policy evidence in the inspected text. This limits what this review can claim; it does not establish that every suggestion is wrong.

One-page owner worksheet for the first route

1 · Approved factsRecord brief ID, owner, audience, factual signup trigger, approved product promise, terms, evidence IDs and exact destination URL. Missing facts stay “not supplied.”

2 · Five draft rowsFor emails 1–5, write purpose, proposed delay, subject, preview, body, one CTA and URL, any segment condition, claim-to-proof link and unresolved question. Unsupported template beats are holds.

3 · Measurement questionsRecord goal-event definition, denominator, window, source, authorized aggregate baseline or “not supplied,” competing emails and journey overlap. Generic source rates are not targets.

4 · Review, then separate releaseRecord named marketing/brand and recipient-policy reviews, proof and URL checks, and specialist checks where required. Keep status “local draft only” until a later explicit ESP/action approval.

This worksheet is pool-proposed, not source-enforced. The owner-maintained template is the first route for one local draft because its fields and hold markers can be controlled directly; that is a design choice, not an observed speed or quality advantage. A third-party skill could be compared later only against repeated, consented omissions on the same approved cases after a full revision, host, data and reviewer check. Neither route was executed here.

What would change the decision

Remove or further narrow the card if a verified intended revision requires recipient data, a live account, ESP configuration or sending to produce this bounded draft, or cannot keep unsupported claims and metrics separate. Consider a controlled template-versus-skill draft comparison only if recurring consented cases reveal specific omissions and named reviewers can inspect both outputs. No such evidence is present.

There was no real audience, consent basis, list, URL, campaign, performance data, local product context, ESP or host installation in this run. No legal, deliverability, conversion, safety, compatibility, quality or performance outcome is established. Run 6, Run 15 and Run 16 remain dated historical observations, not silent 2.1.3 updates.

Current catalogue · Browse by job · Research library · Review a skill