Brief B — Launch Content Kit (executed under Prompt 02 orchestrator + Prompt 03 worker specs)

DEFINITIONS
  "The site" = prompts.mikee.pro as deployed: index, deep-research,
    research, inventory, library (95 prompts), kit (9 files), runs.
    A feature, page, statistic or URL that is not actually on the
    site does not exist for purposes of this kit.
  "Complete email" = subject line + preview text (<=90 chars) +
    body (<=200 words) + one CTA, all present, no placeholders.
  "Complete social post" = platform + copy within that platform's
    limit (LinkedIn 300, X 280, Instagram caption 125 words) + goal.
  "Complete ad" = ≤30-character headline, ≤90-char description1,
    ≤90-char description 2 (Google); ≤150-char primary text
    (LinkedIn), plus a display URL path.
  "Placeholder" = any remaining [bracket], <angle bracket>, TBD,
    "insert here", or "[...]" left in shipped copy.

TASK (goal)
  Four artifact files exist under prompts-site/runs/, each satisfying
  its own count-and-content predicate below, internally consistent
  with each other and with the real site, and each ADMITted by a
  fresh-context adversarial verifier scoring 2 on every criterion.

  ARTIFACT 1 — campaign-brief.md: campaign name + one-sentence
    summary, target audience (primary + secondary), >=3 messaging
    pillars each with one proof point drawn from the actual site
    (page names, real counts like 95 prompts / 9 kit files / 15+
    pages), channel plan, timeline with >=4 milestones, KPIs with
    numeric targets, >=3 risks with mitigations.
  ARTIFACT 2 — email-sequence.md: 5 complete emails (see defs) as
    a cold-to-demo nurture sequence for the site's audience
    (business/marketing operators), each email numbered with its
    funnel job, no placeholders.
  ARTIFACT 3 — social-calendar.md: 14 complete social posts across
    at least LinkedIn, X and Instagram, at least 3 repurposing an
    actual site asset (library, kit, research, runs), within
    platform limits, no placeholders.
  ARTIFACT 4 — ads-landing.md: 10 Google ads (RSA-shaped: 3
    headlines <=30 chars + 2 descriptions <=90 chars each) + 6
    LinkedIn ads (primary <=150 chars + headline <=60 chars), plus
    one full landing-page hero section (headline <=10 words,
    subhead <=30 words, 3 benefit bullets, CTA), no placeholders,
    every claim traceable to the real site.

  The obligation is exact: every count met or exceeded, zero
  placeholders, zero invented site features, and every artifact
  ADMITted on all verifier criteria at grade 2.

  Assume for purposes of this task that a complete solution exists.

DOES NOT COUNT
  - summaries, outlines, or "here is what the kit would contain"
  - copy that contains any placeholder (listed in DEFINITIONS)
  - claims about site features, counts, or URLs that are not on the
    real site (e.g. pricing pages, free trials, customer logos,
    statistics that were not measured)
  - meeting a count by duplicating near-identical items
  - character/word counts asserted rather than actually within
    limits
  - a plan, status report, or "the remaining copy is routine"
  - artifacts that contradict each other (different campaign names,
    audiences, or offers across files)

ORCHESTRATION
  4 concurrent workers, one artifact family each (strategy / email
  / social / paid). Per-spawn: four-part spec from Prompt 03.
  Workers are blind to each other's output; no coordination.
  Registry: one entry per family; all 4 families open, none crowded.
  Synthesis, cross-file consistency, and redirection are the
  orchestrator's job — workers return artifacts only.

VERIFICATION
  Fresh-context adversarial verifier (Prompt 04), separate from the
  orchestrator and workers, given only: the 4 artifacts, this
  predicate, this DOES NOT COUNT list, and the hunt list:
  - invented site features / URLs / statistics (confounder)
  - placeholders surviving in shipped copy (degenerate case)
  - limits asserted rather than met (measurement failure)
  - cross-artifact contradiction (leakage path)
  - CIRCULARITY analogue: a kit that merely describes or promises
    the campaign instead of containing executable copy — "the
    artifact IS the deliverable" cannot be satisfied by narrating
    it.
  Grade A-F criteria 0-2 per Prompt 04. ADMIT only at 2 everywhere.
  Verifier admits/rejects only; repairs route back to workers.

REPORTING CONTRACT
  Every completion claim points at a file with byte size or a
  counted line from this session. Verifier verdicts pasted in full.

RETURN CONDITION
  Return only when all 4 artifacts are ADMITted. Otherwise return
  the strongest verified state plus the exact remaining gap,
  labeled incomplete.

EFFORT
  If a round is rejected, launch a repair round with the REJECT
  reason verbatim; do not return after the first rejection.