L

Knowledge Pack Files

Landing Website Standards Skill Pack Files

Browse the source files that power the Landing Website Standards MCP server knowledge pack.

Available free v1.1.0 LLM
$ sidebutton install landing-website
Download ZIP
best-practices/references/acceptance.md
2.0 KB

Acceptance Bar — AC1–AC5

The gate every generated landing page is judged against — mock first, built page after. Versioned with this skill, not with any run: the design job self-checks and reports a per-check line in its resolution message, QA re-checks independently, and the approval gate reads both. Change this file and every future landing goal is judged by the new bar — edit deliberately.

Binary per check, honestly reported: a failed check is fixed or reported as failed with why — never silently skipped, and "N/A" is not a verdict (a check that cannot apply means the requirements deviated, which is itself worth a line in the report).

IDCheckPass means
AC1Above the fold at 375pxHeadline (~6 words), subhead, primary CTA, and one trust signal visible at 375px width with zero scroll
AC2One conversion actionThe CTA repeats after each major section toward one destination page-wide; no nav, no outbound links; footer legal-only
AC3Narrative orderhero → problem → solution (3 benefits) → proof → FAQ → final CTA, or the deviation the intake decisions recorded; every section has a distinct communication job
AC4Real headlinesThe hero headline and every section heading are real proposed copy; placeholders only in body/media, each labelled with the asset it awaits
AC5Claims traceEvery claim (numbers, outcomes, audience facts, comparisons) traces to an intake input via the copy skeleton's claims-to-source table; reported as a fraction ("9/9 claims trace")
StageWho checksHow
Design — mockdesign job (self-check)All five on the mock + copy skeleton; per-check line in the resolution message
Design — reviewQA (independent)Same five, fresh eyes; diffed against the self-check
Build — pageQAOn the rendered page: AC1 at a real 375px viewport, AC2 by walking every interactive element
Launch reviewoperatorReads both reports; a self-check/re-check disagreement is itself a finding