# zicula-ai-content-qa-kit — free sample

Three real files from the full kit (17 files), each **trimmed to taste** — enough to
judge the quality, not enough to run a defensible QA process:

- `sop-excerpt.md` — the backbone of the pre-publish QA SOP: all four principles
  (multi-evidence only, INCONCLUSIVE by design, a named human decides, the record
  is the product), the complete trigger matrix for when a full check runs, and the
  complete 5-outcome decision table. Steps 1–3 and 5–7 — the file-freeze fields,
  the AI-SUSPECTED conversation script and the INCONCLUSIVE protocol that resolves
  on draft history instead of another score — are summarized at the end.
- `consensus-excerpt.md` — the anti-score-as-proof core of the detector checklist,
  complete: Rule zero (official APIs only), the three-sentence case for consensus,
  all four panel-selection criteria, and the mandatory non-English / non-native
  default (INCONCLUSIVE, per the Stanford 61.22% false-flag study). The four-vendor
  panel table is cut to 2 of 4 rows; the run checklist, quarterly re-check and
  credit math are summarized.
- `disclosure-excerpt.md` — the complete English disclosure statements for Tier 1
  (assisted) and Tier 2 (generated), short and long forms, plus the complete
  4-option placement guide and the "what a disclosure must not say" rule. The
  statements in ES / PT-BR / DE / JA / TH and the two client-email templates are
  in the full kit.

Every excerpt is honest about what was cut — the cuts are listed at the end of each
file, so you know exactly what the full kit adds.

What the full kit adds (the other 14 files, plus the uncut versions of these three):

- The 3-tier AI-use policy: content-type matrix (where "generated" is simply not
  accepted), writer and editor obligations, client-facing summary for your MSA,
  enforcement steps
- The full QA SOP: 7 steps from file freeze to sign-off, with the sign-off sentence
  format and the quarterly review routine
- The honest-limitations brief: primary sources on what detectors can't do
  (OpenAI classifier retirement, Vanderbilt/Turnitin, Stanford) — the receipts
  behind the kit's design
- The evidence-pack CSV (31 columns, 4 worked example rows) and risk-log CSV, plus
  a small audit script that blocks incomplete records from being archived
- The full consensus checklist: all four vendor rows with date-stamped pricing
  (captured 2026-09-20), the run checklist, and the credit math
- The style-guide insert for your team handbook — English in full, core rules in
  five more languages
- Five screenshots previewing the quickstart, the QA SOP, the consensus
  checklist, the AI-use policy and the disclosure statements

One-time purchase $42 · free updates for v1.x · 30-day full refund.