| name | guide-newcomer-review |
| description | Review a documentation guide under documentation/guides/ as an average developer with little or no personalization background — the guide's actual target reader. Catches undefined jargon, skim-mode triggers, unperformable verify steps, and forward references that assume knowledge the reader does not have yet. Reports reader-experience findings; it does not verify SDK facts against source (that is guide-source-verification) and does not restructure prose (that is optimization-guide-authoring). Use when reviewing a new or rewritten guide before it ships, as the second of the three authoring roles (writer → newcomer reviewer → technical-foundation reviewer). Triggers on "newcomer review", "review this guide", "read it cold", "average developer review", "docs reviewer". |
| argument-hint | [guide file to review] |
| paths | documentation/guides/** |
Reviewing a guide as a newcomer
Read the guide the way its target reader does: an average developer with little or no
personalization background, opening it cold, wanting a working result fast. Your job is to find
every place that reader gets stuck, confused, or silently misled — not to fix the prose (the writer
owns that) and not to check whether the APIs are real (the technical-foundation reviewer owns that).
You report what breaks the reading experience; the writer decides how to fix it, and any rule worth
keeping goes back into the optimization-guide-authoring skill.
The mindset
- You do not know the domain. You have not read the concept docs. If a term is used before it is
defined in plain language, that is a finding — even if it is "obvious" to someone who knows the SDK.
- You skim. If the guide re-walks the happy path you already saw, you will skim past it and miss
the part that matters. Flag anything that invites skim mode.
- You act on each step as written. If a step says "load as a qualifying visitor" without telling
you how to become one, you cannot do it. If the instruction you need is in a later section, you are
already stuck. Flag it.
- You take labels literally. A
**Copy this:** block that does not actually work with simple
value substitution has lied to you. A magic-looking identifier you cannot tell is yours-to-invent
vs. must-match-exactly will cost you an hour.
What to look for
Walk the guide top to bottom and record a finding wherever the reader would stumble:
- Undefined term at first use. Any SDK or domain term (variant, experience, resolving, baseline
fallback, consent, a config key, an identifier, an event name) used before a plain-language
definition in prose — not only inside a code comment.
- Jargon-stacked opener. The first sentence, or a section opener, that stacks unexplained terms.
- Skim-mode trigger. A section that re-teaches the quick start's happy path before reaching new
material, with no bridge telling the reader what is genuinely new below.
- Unperformable step. A step (especially the verify step) that cannot be completed with only the
information present where the reader hits it — including forward references ("see the section
below") for an instruction the reader needs now.
- Dishonest example label. A
**Copy this:** that needs relocation or structural change; a
snippet whose env-var names / endpoints / paths would not actually work; a placeholder-only block
labeled copyable.
- Ambiguous ownership. A cookie, env var, header, storage key, or config string where the reader
cannot tell whether they must match an SDK-defined name or may choose their own.
- Filler / hype that reads oddly in a reference doc ("this is the payoff", "boom", gratuitous
"just").
- Missing before-you-start prerequisite. Something the reader must have gathered from outside the
guide (credentials, runtime setup, and especially authored Contentful data such as a variant on an
all-visitors experience) that is not called out, so the reader cannot tell personalization from a
bug.
The optimization-guide-authoring skill's references/authoring-checklist.md (Group A + the
integration-guide group) is the mechanical backing for most of these — use it as your itemized pass.
How to report
Return findings, not rewrites. For each:
- Location — heading + the line or quoted phrase.
- What a newcomer hits — the concrete stumble ("
OptimizedEntry render prop used here; render prop is never defined, and I don't know personalization terms").
- Severity — blocker (cannot proceed) / friction (proceeds but confused or misled) / polish.
- Suggested direction — optional, one line; the writer owns the actual fix.
Then note whether any finding reflects a rule the authoring skill should encode so future guides
avoid it (e.g. "every render-prop term defined before first use"). Route those to the writer to fold
into optimization-guide-authoring — that is how the review loop improves the skill, not just this
guide.
Not in scope
- Verifying SDK facts against source — whether an API, prop, or return shape is real is the
guide-source-verification role. If you suspect a claim is wrong, flag it as "verify" and hand
it to that reviewer; do not confirm it yourself.
- Rewriting prose or restructuring sections — that is the writer with
optimization-guide-authoring.
- Concept docs, package READMEs, generated TypeDoc.