ceo-orchestration 1.0.0

package/.claude/skills/domains/marketing-global/skills/book-co-author/SKILL.md

@@ -0,0 +1,220 @@
+---
+name: book-co-author
+description: >
+  Ghostwriting and co-authorship discipline covering voice capture, narrative
+  architecture, chapter structure, source management, attribution clarity, and
+  the manuscript-to-published lifecycle. Distinct from blog or campaign content
+  by scope, voice persistence requirement, and contractual co-authorship
+  structure. Applies premise-promise-payoff as the primary manuscript frame,
+  enforces interview-based voice discovery over draft-submission, and gates all
+  editorial decisions on developmental sequence before copy or proof passes.
+  Use when: ghostwriting or co-authoring a full-length nonfiction manuscript;
+  establishing a voice document for a long-form project; structuring a chapter
+  outline before prose drafting; resolving attribution or rights language at
+  contract stage; selecting between traditional, hybrid, or self-publishing
+  paths; or auditing a manuscript-in-progress for voice drift, claim sourcing
+  gaps, or skipped editorial sequence.
+owner: Valentina Sørensen (Book Co-Author, domain persona)
+tier: domain:marketing-global
+scope_tags: [book-authoring, ghostwriting, voice-capture, manuscript-architecture, publishing, co-authorship]
+inspired_by:
+  - source: msitarzewski/agency-agents/marketing/marketing-book-co-author.md@783f6a72bfd7f3135700ac273c619d92821b419a
+    license: MIT
+    relationship: structural_inspiration
+    authored_by: ceo-orchestration framework
+    authored_at: 2026-05-07
+# --- smart-loading fields (PLAN-083 Wave 0b sub-agent 0.7c) ---
+domain: marketing-global
+priority: 8
+risk_class: low
+stack: []
+context_budget_tokens: 500
+inactive_but_retained: true
+repo_profile_binding:
+  frontend: {active: false, priority: 10}
+  engine: {active: false, priority: 10}
+  fintech: {active: false, priority: 10}
+  trading-readonly: {active: false, priority: 10}
+  generic: {active: false, priority: 10}
+activation_triggers: []
+# --- K1 paths: native file-touch activation (PLAN-135 W3 unit k1a) ---
+paths:
+  - "**/manuscript/**"
+  - "**/chapters/**"
+  - "**/book/**"
+---
+
+# Book Co-Author
+
+## Cardinal Rule
+
+Co-authorship without explicit attribution language at contract signing is a
+future legal dispute waiting to be filed. Ghost, co-author, "with" credit, and
+acknowledgment acknowledgment are four distinct legal positions with different
+royalty, rights, and subsequent-works implications. Resolving them after
+manuscript delivery is expensive and adversarial. Establish attribution
+structure, royalty split, and rights to derivative works in the contract — not
+the preface.
+
+## Fail-Fast Rule
+
+Refuse to proceed to prose drafting if any of the following are unresolved:
+no voice document exists; chapter premise is not a falsifiable claim; source
+attribution for reproduced material is absent; or the author's name-and-credit
+arrangement has not been stated in writing. Unresolved, these items corrupt
+every page that follows.
+
+## When to Apply
+
+Apply when the engagement scope is a full manuscript of 30,000 words or more,
+or a defined section thereof with a discrete chapter promise. Do not apply to
+blog posts, long-form articles under 5,000 words, or social serialisation
+projects — those belong to `content-creator`. Apply when voice persistence
+across twelve or more chapter drafts is a contractual deliverable, not a
+courtesy.
+
+## Voice Capture Discipline
+
+Interview transcripts are the primary source for voice discovery. Written
+drafts submitted by the author reflect their edited self-image, not their
+natural register — transcripts expose idiom, cadence, sentence rhythm, and
+signature phrases that edited prose suppresses.
+
+Produce a voice document before any prose is drafted. The voice document
+records: dominant sentence length and subordinate-clause frequency; recurring
+idioms and field-specific vocabulary; stance markers (hedging vs. declarative
+preference); topics the author avoids or escalates; and three to five
+representative passage excerpts annotated for register.
+
+Run a voice-drift check at the end of every chapter by comparing a 500-word
+passage against the voice document. Flag passages where sentence rhythm,
+vocabulary register, or stance shifts without a structural reason. Voice drift
+is not a style preference — it signals the author's credibility is eroding.
+
+## Manuscript Architecture
+
+Structure the manuscript around a premise-promise-payoff frame. The premise
+is the problem or tension the book resolves. The promise is the specific
+transformation the reader will undergo. The payoff is the evidence or
+argument that closes the loop. All three must be stated in the book proposal
+before chapter outlines are written.
+
+Produce a chapter outline before drafting prose. The outline for each chapter
+states: the chapter promise (one sentence), the core argument, the supporting
+evidence or scene, the reader takeaway, and the strategic role within the full
+manuscript arc. Chapters that serve no strategic role are cut at outline
+stage, not after drafting.
+
+Enforce one dominant idea per chapter. Chapters that attempt to resolve two
+unrelated arguments produce reader confusion and inflate word count without
+improving argument quality. Split or cut before drafting.
+
+Include a reader-takeaway statement per major section. The takeaway is a
+single declarative sentence the reader should be able to reproduce after
+reading the section. If the takeaway cannot be stated in one sentence, the
+section has not resolved its argument.
+
+## Source Management
+
+Archive all interview audio and transcripts with date, speaker, and session
+identifier. Transcripts are primary sources — treat them with the same
+provenance discipline applied to published references.
+
+Record quoted-source provenance for every claim that will appear in print:
+author, title, publication, year, and page or URL. A claim without provenance
+is a liability at copy-edit stage and a credibility failure after publication.
+
+Obtain written permissions for any reproduced material before the manuscript
+enters copy-edit. Verbal permissions are not permissions. Permissions obtained
+after copy-edit delay publication and increase legal exposure.
+
+Do not invent or paraphrase quotes. If an interviewee's recorded statement
+does not support the argument cleanly, either find a statement that does or
+reframe the argument. Constructed quotes that approximate what a source
+"probably said" are a publication-ending error.
+
+## Attribution Discipline
+
+Four credit structures exist for collaborative book projects: ghost (no public
+credit for collaborator), co-author (shared byline), "with" credit (named
+secondary contributor on cover), and acknowledgment (named in front matter
+only). Each carries different royalty structures, rights to derivative works,
+and public-identity implications.
+
+Define at contract stage: which credit structure applies; the royalty split
+and advance allocation; whether the collaborator retains any rights to the
+manuscript content; and whether the collaborator may reference the project in
+their own portfolio. Defining these terms after manuscript delivery introduces
+leverage asymmetry and is the most common source of post-project disputes in
+ghostwriting engagements.
+
+## Editorial Cycle
+
+The editorial sequence is: developmental edit → line edit → copy edit →
+proof. This sequence is not a preference — it is a dependency chain.
+Developmental edit resolves argument structure, narrative coherence, and
+chapter-level logic. Line edit resolves prose rhythm, paragraph transitions,
+and sentence-level clarity. Copy edit resolves grammar, consistency, and
+style-guide conformance. Proof resolves typographical and layout errors.
+
+Skipping developmental edit under time pressure is the most common editorial
+error in first-time book projects. Developmental problems discovered at copy
+edit are more expensive to fix than at developmental stage, and developmental
+problems discovered post-publication are not fixable. If the schedule does
+not accommodate a developmental pass, the schedule is wrong — not the
+editorial sequence.
+
+## Publishing Path Selection
+
+Three primary publishing paths exist: traditional, hybrid, and self-publishing.
+
+Traditional publishing offers the strongest prestige signal, widest
+distribution infrastructure, and zero author-side production cost, at the
+cost of loss of creative control, 12-to-24-month timelines post-acquisition,
+and royalty rates of 8 to 15 percent of net. Suitable when the author's
+primary goal is category authority in a credentialed professional domain.
+
+Hybrid publishing offers faster timelines (six to twelve months) and
+greater creative control than traditional, at the cost of author-borne
+production costs of $5,000 to $50,000 and reduced distribution reach. Suitable
+when speed and control outweigh prestige and the author has a direct audience
+to sell into.
+
+Self-publishing offers full creative control and maximum royalty rates, at
+the cost of full production-cost burden, no prestige signal from a named
+imprint, and distribution limited to platforms the author manages directly.
+Suitable for authors with established direct audiences and no need for
+third-party credentialing.
+
+Select the path before manuscript completion — not after. Path selection
+determines contract structure, cover design constraints, distribution
+registration timelines, and ISBN assignment. Reversing path after manuscript
+completion wastes production spend.
+
+## Anti-patterns
+
+| Anti-pattern | Consequence | Mitigation |
+|---|---|---|
+| Stylistic drift across chapters | Author's voice becomes unrecognisable by chapter six; readers perceive multiple authors | Voice document + end-of-chapter voice-drift check |
+| Source fabrication | Legal liability and publication retraction | Provenance record for every claim; no invented quotes |
+| Attribution ambiguity at contract | Post-delivery dispute over royalties and rights | Attribution structure defined in writing before work begins |
+| Skipped developmental edit | Structural argument failures discovered at copy edit or post-publication | Enforce editorial sequence; reject schedule that omits developmental pass |
+| Unsigned reproduction permissions | Publication delay or legal action from rights holders | Obtain written permissions before copy-edit stage |
+| Voice document from drafts only | Voice document captures edited self-image, not natural register | Interview transcripts as primary source; annotate for cadence and idiom |
+| Premise not falsifiable | Chapter cannot be argued or defended; degenerates into opinion | State premise as a claim that can be contested before outline is written |
+
+## Cross-References
+
+- `domains/marketing-global/skills/content-creator` — long-form articles and
+  repurposing below book scope
+- `domains/marketing-global/skills/podcast-strategist` — interview-based
+  content aligned to voice capture methodology
+- `domains/business-support/skills/legal-compliance-checker` — contract review
+  for attribution and rights language
+
+## ADR Anchors
+
+- ADR-058: creative-rewrite threshold (≤12-word verbatim; ≤4-word H2; full
+  independence from upstream prose)
+- ADR-060: bulk creative authoring strategy — domain skill authorship policy
+  and structural-inspiration relationship classification

package/.claude/skills/domains/marketing-global/skills/carousel-growth-engine/SKILL.md

@@ -0,0 +1,393 @@
+---
+name: carousel-growth-engine
+description: >
+  Cross-platform carousel post engineering — slide architecture, visual
+  hierarchy, hook-to-payoff arc, and save-share mechanics across Instagram,
+  LinkedIn, Twitter (X carousel / document), TikTok photo series, and
+  Pinterest. Covers the 10–12 slide ceiling, one-idea-per-slide discipline,
+  cover-slide-as-ad design, and the asymmetric value of saves versus likes
+  in algorithm distribution. Use when: designing a carousel from scratch;
+  auditing a carousel for structural or narrative weaknesses; adapting an
+  existing carousel to a new platform's aspect ratio and character limits;
+  running A/B tests on cover or payoff variants; or diagnosing low save
+  rates on formats expected to produce reference-worthy content.
+owner: Rafael Sousa (Carousel Strategist, domain persona)
+tier: domain:marketing-global
+scope_tags: [carousel-design, slide-architecture, visual-hierarchy, save-economy, multi-platform-carousel, document-format]
+inspired_by:
+  - source: msitarzewski/agency-agents/marketing/marketing-carousel-growth-engine.md@783f6a72bfd7f3135700ac273c619d92821b419a
+    license: MIT
+    relationship: structural_inspiration
+    authored_by: ceo-orchestration framework
+    authored_at: 2026-05-07
+# --- smart-loading fields (PLAN-083 Wave 0b sub-agent 0.7c) ---
+domain: marketing-global
+priority: 8
+risk_class: low
+stack: []
+context_budget_tokens: 500
+inactive_but_retained: true
+repo_profile_binding:
+  frontend: {active: false, priority: 10}
+  engine: {active: false, priority: 10}
+  fintech: {active: false, priority: 10}
+  trading-readonly: {active: false, priority: 10}
+  generic: {active: false, priority: 10}
+activation_triggers: []
+# --- K1 paths: native file-touch activation (PLAN-135 W3 unit k1a) ---
+paths:
+  - "**/carousels/**"
+  - "**/slides/**"
+---
+
+# Carousel Growth Engine
+
+## Cardinal Rule
+
+A carousel that does not earn the swipe at slide 1 reaches 0% completion
+regardless of payoff quality. The cover slide is not an introduction — it is
+an advertisement for the rest of the deck. Its function is to create an
+information gap specific enough that closing it requires swiping. A cover that
+restates the topic without creating tension, or that frames the payoff as
+optional ("you might find this useful"), will not produce swipe events at
+scale. The entire save-share economy that follows slide 1 is contingent on
+that first swipe. Design the cover before designing the deck.
+
+---
+
+## Fail-Fast Rule
+
+A carousel MUST NOT be published if any of the following conditions hold:
+
+1. The cover slide does not contain a hook that identifies a specific problem,
+   claim, or information gap — generic topic labels ("5 things about X") are
+   insufficient; the tension must be implicit or explicit.
+2. Any individual slide contains more than one distinct idea; a slide that
+   serves two purposes serves neither.
+3. The total slide count exceeds 12; completion rate degrades non-linearly
+   above 10 slides, and the payoff slide is deprioritized when preceded by
+   cognitive overload.
+4. The payoff slide (final slide or penultimate slide before CTA) delivers no
+   concrete action, insight, or resolution — a summary of what was just shown
+   is not a payoff.
+5. The deck has not been validated against the target platform's aspect ratio
+   before export; mismatched dimensions produce cropped text and broken visual
+   hierarchy at render time.
+
+Halt production and resolve before scheduling.
+
+---
+
+## When to Apply
+
+Activate this skill when the work involves any of the following:
+
+- Designing a carousel post for Instagram, LinkedIn, Twitter (X), TikTok
+  photo series, or Pinterest — from brief to final slide export.
+- Auditing an existing carousel for narrative arc integrity, visual hierarchy
+  failures, or platform-native format compliance.
+- Adapting a carousel built for one platform to the dimension and copy
+  constraints of another.
+- Diagnosing low save rates or low completion rates on carousel formats that
+  are expected to function as reference-worthy content.
+- Running cover-slide or payoff-slide A/B variants to identify which hook
+  framing or CTA structure performs best for a given audience.
+- Establishing a reusable carousel template system with locked grid, type
+  hierarchy, and color system for a brand or creator.
+
+Skip when: the task is a single static image (no slide sequence involved);
+the task is a Reels or short-form video (use
+`domains/marketing-global/skills/tiktok-strategist` or
+`domains/marketing-global/skills/instagram-curator`); or the task requires
+paid carousel ad creative (paid format mechanics differ from organic carousel
+distribution).
+
+---
+
+## Slide Architecture
+
+### Ceiling and Density
+
+The 10–12 slide ceiling is a completion-rate constraint, not a creativity
+constraint. Platform analytics consistently show completion drop-off beginning
+at slide 8 for casual discovery contexts and at slide 11–12 for high-intent
+educational content. The ceiling is 12 as an absolute maximum; 7–10 slides
+is the functional target range for most carousel types.
+
+One idea per slide is non-negotiable. A slide that contains two parallel
+points, two data references, or two sub-claims forces a reading hierarchy that
+mobile viewers at scroll speed will not resolve. Split the slide. If splitting
+produces too many slides, the architecture problem is upstream — the ideas
+themselves require restructuring, not the slide layout.
+
+### Cover Slide as Advertisement
+
+The cover slide performs the function of a paid ad creative: it must stop the
+scroll, communicate the value proposition, and create a desire for the
+next frame — all within the first 300 ms of processing. Design constraints
+that follow from this function:
+
+- The hook text must be visible without zooming; headline type at a minimum
+  equivalent to 32px at 1080×1080.
+- The visual element (image, illustration, or typographic treatment) must
+  support the hook, not compete with it; decorative background that reduces
+  text contrast degrades performance.
+- The cover slide must work as a standalone unit when displayed in a feed
+  thumbnail or link preview — assume the viewer will only ever see slide 1.
+
+### TOC Slides as Roadmap
+
+For carousels above 7 slides targeting educational or reference content, a
+slide 2 table of contents anchors the reader's expectation and reduces early
+abandonment. The TOC slide should enumerate the payoffs the viewer will
+receive, not the topics that will be covered — framed as outcomes, not
+subjects.
+
+### Payoff Slide as CTA
+
+The payoff slide (typically the final content slide, followed optionally by
+a CTA-only slide) must deliver the most concentrated value in the deck. It
+should be designed as the slide most likely to be saved — a framework,
+checklist, decision tree, or key takeaway that earns return reference. The
+CTA that follows should direct to a single action, not a menu of options;
+multiple CTAs on a closing slide split intent and reduce conversion.
+
+---
+
+## Hook-Development-Payoff Arc
+
+### Cover Hook: Thesis and Curiosity
+
+The cover hook operates on one of three mechanics: tension (a claim the viewer
+wants to verify), pain point (a problem the viewer recognizes), or curiosity
+gap (a pattern revealed partially, requiring the deck to complete it). Each
+mechanic works; mixing more than one mechanic on a single cover weakens both.
+
+A thesis hook states a position the viewer either agrees with strongly enough
+to validate via swipe, or disagrees with strongly enough to contest via swipe.
+Neutral statements do not produce either response. A curiosity-gap hook names
+the gap explicitly and withholds the resolution. A pain-point hook names the
+problem with enough specificity that the target audience recognizes it as
+their own — generic pain points ("feeling overwhelmed") produce lower swipe
+rates than specific ones ("spending 3 hours on content that gets 12 likes").
+
+### Development Slides: Stepping Stones
+
+Development slides (slides 2 through N−1) function as stepping stones from
+the hook to the payoff. Each slide advances the argument by one unit — a data
+point, a principle, an example, a counter-argument addressed — without
+restating what the previous slide established. The development section fails
+when slides begin repeating information in paraphrased form to fill the slide
+count. If the argument has 5 steps, the deck has 5 development slides plus
+cover and payoff. Do not inflate.
+
+Formatting discipline within development slides: use one visual anchor per
+slide (a single statistic, a labeled diagram, a short list of three items or
+fewer). Text-heavy slides that require reading rather than scanning produce
+completion drop-off regardless of content quality.
+
+### Payoff: Action or Insight
+
+The payoff slide provides either a concrete action (do this next, use this
+framework, apply this decision rule) or a novel insight (the reason behind
+the pattern, the principle that unifies the development slides, the
+implication the viewer had not considered). A payoff that summarizes the
+development is not a payoff — it is a recap, and it does not earn a save.
+
+---
+
+## Visual Hierarchy
+
+### Typography Hierarchy
+
+Every slide must establish a clear reading order through type size, weight,
+and positioning. The primary message (headline or key statement) should be
+processable in isolation — a viewer who reads only the headline of each slide
+should be able to follow the arc of the argument. Supporting text provides
+evidence or elaboration but is secondary in visual weight.
+
+Hierarchy levels for carousel slides:
+- Level 1: headline or key claim — largest size, highest contrast, positioned
+  at optical center or top third.
+- Level 2: supporting statement or data label — 60–70% of Level 1 size;
+  subdued weight or reduced opacity.
+- Level 3: attribution, footnote, or secondary context — minimal size, lowest
+  contrast, positioned at bottom margin.
+
+Never use more than two typefaces within a carousel. Typeface consistency
+across all slides is a brand-recognition signal at thumbnail scale.
+
+### Contrast and Legibility
+
+Contrast ratio between text and background must meet a minimum of 4.5:1
+(WCAG AA) for body text and 3:1 for large display text. This is a baseline,
+not a creative floor — many carousel platforms render under variable ambient
+light conditions; a contrast ratio of 7:1+ is the functional target for mobile
+contexts. Test all text overlays on image backgrounds; image carousels with
+photography backgrounds require a scrim or text panel to ensure legibility.
+
+### Negative Space
+
+Negative space is structural, not stylistic. A slide with zero negative space
+compresses the visual hierarchy and eliminates the resting points the eye uses
+to process content at scroll speed. Maintain a minimum margin of 8–10% of
+slide width on all sides. Key information should not bleed to within 5% of
+any slide edge — platform overlays, swipe indicators, and profile elements
+encroach on the edges at render time.
+
+### Consistent Grid
+
+A grid system applied consistently across all slides in a carousel signals
+production craft and reinforces brand identity. Define column structure, row
+baselines, and margin constants before designing individual slides. All text
+anchors, image crops, and iconographic elements should align to the grid.
+Inconsistent element placement between slides creates visual noise at swipe
+speed and undermines the coherence the carousel architecture requires.
+
+---
+
+## Platform-Native Variants
+
+Each platform applies different aspect ratios, character limits, and rendering
+behaviors. A carousel designed for one platform and republished without
+adaptation will render with cropped text, collapsed captions, or broken visual
+hierarchy on others.
+
+| Platform | Aspect Ratio | Optimal Resolution | Caption Limit | Notes |
+|---|---|---|---|---|
+| Instagram | 1:1 or 4:5 | 1080×1080 or 1080×1350 | 2,200 chars | 4:5 crops more feed real estate; first slide most critical for save |
+| LinkedIn | 4:5 or 1:1.294 | 1080×1350 or 1080×1400 | 3,000 chars | Document (PDF) format supported; LinkedIn carousels indexed as documents |
+| Twitter (X) | 16:9 or 3:4 | 1200×675 or 900×1200 | 280 chars per tweet | Tweet thread with attached images; each tweet is its own unit |
+| TikTok Photo | 9:16 | 1080×1920 | 2,200 chars | Photo series format; vertical full-bleed; no text in bottom 20% |
+| Pinterest | 2:3 | 1000×1500 | 500 chars | Idea Pins support up to 20 pages; each page = one slide equivalent |
+
+Platform-specific design constraints beyond dimensions:
+
+- **Instagram**: save behavior is the primary signal; design for reference
+  reusability, not one-time consumption.
+- **LinkedIn**: document carousel format (PDF upload) enables download as
+  a signal — high-value reference content earns document saves distinct from
+  regular engagement.
+- **Twitter (X)**: each slide must read as a standalone tweet image; the
+  thread arc supplements but cannot substitute for per-slide independence.
+- **TikTok**: no text within the bottom 20% of slide height — TikTok UI
+  overlays (like/comment buttons, caption) obscure this zone.
+- **Pinterest**: Idea Pins are indexed by Pinterest search; include keyword-
+  rich text overlays for searchability in addition to aesthetic design.
+
+---
+
+## Save-Share Economics
+
+### Save: "I Will Need This Later"
+
+A save is a forward-looking behavior — the viewer identifies the content as
+having future reference value and stores it for retrieval. Save-oriented
+design means the carousel must function as a reference artifact: structured,
+scannable, and dense with actionable information. Formats with the highest
+intrinsic save rate: checklists, decision frameworks, comparison tables,
+step-by-step processes with labeled steps, and curated resource lists.
+
+The payoff slide is the save trigger. If the payoff slide delivers a framework
+or decision tool dense enough to warrant re-consultation, save rate will reflect
+it. If the payoff slide is a motivational statement or brand recall message,
+save rate will be suppressed regardless of development quality.
+
+### Share: "They Need This Now"
+
+A share is a relationship-directed behavior — the viewer identifies the content
+as relevant to a specific person or network and routes it there. Share-oriented
+design means the carousel must contain a slide or statement the viewer can
+use as a social signal: something that articulates their experience, validates
+their position, or provides information that makes them appear informed or
+helpful to their network.
+
+Shares to Stories (Instagram, Facebook) extend organic reach beyond followers
+without algorithm penalty. Shares via DM indicate high personal relevance.
+Design at least one development slide with share-trigger intent — a counter-
+intuitive framing, a specific data point, or a statement that validates a
+professional identity the target audience holds.
+
+### Designing for Both Simultaneously
+
+Save and share are compatible objectives but serve different slide positions.
+The payoff slide earns saves; middle development slides earn shares. A carousel
+designed to optimize only for saves (dense, reference-heavy throughout) tends
+to be too information-dense for share behavior. Design the development arc for
+share triggers at slides 3–5; design the payoff for save retention. A carousel
+that earns both signals receives multiplicative algorithmic distribution.
+
+---
+
+## Carousel A/B Testing
+
+### Cover Variant Testing
+
+Cover slide variants are the highest-leverage A/B test in carousel
+optimization. A 10% improvement in swipe-through rate on the cover compounds
+across all downstream slides. Test one variable per cover variant: hook
+framing (tension vs. pain point vs. curiosity gap), visual treatment
+(typographic-only vs. image-backed), or subject line length (long-form
+headline vs. short punchy claim). Do not change multiple variables
+simultaneously — attribution becomes impossible.
+
+Minimum sample size before declaring a winner: 500 impressions per variant.
+Below this threshold, engagement rate variance from audience sampling dominates
+signal from the actual content difference.
+
+### Payoff Slide Variants
+
+Payoff slide testing identifies whether a framework, checklist, or key insight
+produces higher save rates than a motivational CTA or narrative resolution.
+Run payoff variants with cover held constant. Metrics: save rate (saves / reach)
+is the primary signal; share rate is secondary. Do not use like rate as a proxy
+for payoff effectiveness — likes do not indicate reference intent.
+
+### Same Content, Multiple Aesthetics
+
+Once a carousel narrative arc is validated (the argument structure earns
+swipes and saves), test aesthetic variants with identical copy to identify
+whether the visual system drives incremental performance. Aesthetic testing is
+lower priority than hook and payoff testing — narrative structure has higher
+variance contribution to outcome than visual treatment for most informational
+carousels.
+
+---
+
+## Anti-Patterns
+
+| Anti-Pattern | Why It Fails |
+|---|---|
+| Text-wall slides with multiple paragraphs per slide | Scroll-speed processing cannot resolve multi-paragraph slides; completion rate collapses at first wall slide |
+| Deck with no payoff slide — all content ends at the last development point | No concrete takeaway means no save trigger; the carousel is consumed and discarded |
+| Repetition slides that paraphrase previous slides to fill slide count | Inflated slide count degrades completion rate; repetition signals low information density to the viewer |
+| Decorative visual noise that reduces text contrast | Ornamental backgrounds or patterns that reduce contrast to below 4.5:1 make text illegible at thumbnail scale |
+| Ignoring platform aspect ratio on export | Cropped text and broken visual hierarchy at render time; cover slide hook may be partially hidden |
+| Weak cover hook — topic label instead of tension | "5 tips for productivity" is not a hook; it is a topic; it does not create the information gap that earns a swipe |
+| Multiple CTAs on the closing slide | Split CTA intent reduces conversion; one action per closing slide |
+| Designing development slides for reading, not scanning | Slides requiring 10+ second read time produce early abandonment; one visual anchor per slide maximum |
+
+---
+
+## Cross-References
+
+- `domains/marketing-global/skills/instagram-curator` — Instagram-specific
+  save-engine mechanics, format mix, and aesthetic coherence discipline;
+  shares carousel design overlap but extends to Reels, Stories, and Feed.
+- `domains/marketing-global/skills/linkedin-content-creator` — LinkedIn
+  document carousel and thought-leadership content architecture; covers
+  professional-positioning constraints specific to the LinkedIn algorithm.
+- `domains/marketing-global/skills/content-creator` — cross-platform
+  content creation discipline covering narrative architecture, audience
+  research, and voice consistency; provides the upstream authoring frame
+  within which carousel slide copy is produced.
+
+---
+
+## ADR Anchors
+
+- **ADR-058** — Creative content authoring policy: house-voice rules
+  (declarative prose, no emojis in framework artifacts, no 2nd-person
+  address), maximum verbatim-match threshold with upstream sources, and
+  structural inspiration attribution requirements. All content in this
+  SKILL.md is original prose authored under ADR-058 constraints.