Social Post

Designs social-platform-native visuals (Feed, Story, Reel, Carousel) across Instagram, TikTok, Facebook, Pinterest, RedNote, LinkedIn, X/Twitter and other social platforms by combining nine canonical formats, seven industry visual-DNA presets, a seven-step workflow, a carousel framework with cross-slide consistency anchoring, and a self-audit gate. The skill produces a design description, optional caption, and self-check score per output.

The format taxonomy (Feed Square 1:1 / Feed Portrait 4:5 / Story 9:16 / Reel Cover 9:16 / Carousel 4:5 / etc.) maps directly across Instagram, TikTok, Facebook, RedNote, and Pinterest — the same aspect ratios are platform-native on all of them. Platform-specific safe-zone tuning lives in references/hard-constraints.md.

Engagement Principles

These rules apply across every Section. Read before acting.

1. Lock dimensions first — every request resolves to one of nine formats with explicit width × height and aspect ratio. No format = no work.
2. Industry DNA before composition — match the user's industry to one of the seven presets; load its prompt patch + composition + color + lighting + common errors as a single bundle.
3. Layered constraint priority — Hard Constraints (Section 0) > Industry Rules (Section 1) > User Preferences > Model defaults. When user color overrides, industry default yields.
4. Carousel consistency is per-slide repetition — the consistency anchor is repeated verbatim at the start of every slide's prompt, not declared once.
5. Mandatory self-audit before delivery — every output passes the 14-item check (Section 4) and emits the seven-line Self-Check score.
6. Banned words are filtered out of generator prompts — platform names / UI terms / screenshot terms / brand-tool names never reach the image model. They are used only for internal format selection and industry matching.
7. Match the user's language — respond in English if user writes English, in Chinese if user writes Chinese. Never switch language unprompted.
8. Use <suggestion> for option sets — when proposing platform / format / industry choices, wrap them in <suggestion> tags. Never ask multiple questions at once.

Execution Procedure

generate_social_visual(user_request) → design_description + caption + self_check + image

# Step 0 — Pin hard constraints (MUST, before any decision)
load references/hard-constraints.md
    → Dimensions table / Safe Zones / Text Overlay Rules / Design Quality Baseline / Banned Words
keep these in working context for Steps 1, 5, 6 — violation produces fake UI artifacts,
watermarks, occluded CTAs, or platform-policy breaks.
preflight = enforce_constraints(format=null, output_text=null, prompt_text=user_request)
                                              # references/hard-constraints.md
                                              # pre-flight banned-word scan on the user request
assert preflight.pass

# Step 1 — Confirm Format and Scenario (lock dimensions first)
format = resolve_format(user_request)         # references/hard-constraints.md §Default Format Selection
    if user specified format → use it
    else apply default scenario logic:
        single image content       → Feed Portrait 4:5
        multi-page knowledge       → Carousel 4:5
        short-lived / ad           → Story 9:16
        short-video cover          → Reel Cover 9:16
dimensions, ratio, safe_zone = lookup(format) # references/hard-constraints.md

# Step 2 — Identify Industry (auto-load Visual DNA)
industry_bundle = load_industry_dna(user_request)
                                              # references/industry-dna.md
                                              # returns {prompt_patch, composition, color,
                                              # lighting, common_errors, ig_references}
                                              # on no_match → closest industry as base + adjust

# Step 3 — Define Engagement Goal
goal = SHARE | SAVE | COMMENT
    if user specified → use it
    else infer from content type:
        education / how-to     → SAVE
        lifestyle / humor      → SHARE
        opinion / debate       → COMMENT

# Step 4 — Define the Hook (3-Second Rule)
hook_type = Bold Claim | Curiosity Gap | Visual Shock
    drives Slide-1 design (Carousel) or headline strategy (single image)

# Step 5 — Composition + Color (Industry × Format cross-decision)
composition = industry_bundle.composition + safe_zone constraints
color       = brand_color (if exists, overrides) or industry_bundle.color
lighting    = industry_bundle.lighting
text        = format.text_overlay_rule (references/hard-constraints.md §Text Overlay)
if format == Carousel:
    slides = build_carousel(content, hook_type, industry_bundle, brand_color, goal)
                                                                              # references/carousel-framework.md

# Step 6 — Self-Check Gate (mandatory)
audited_output = run_workflow(user_request)
                                              # references/workflow-and-self-check.md
                                              # runs Section 2 steps + Section 4 14-item audit
findings = enforce_constraints(format, audited_output.output_text, audited_output.prompt_text)
                                              # references/hard-constraints.md
                                              # re-validates sizing, safe zone, overlay, quality, banned words
assert findings.pass
    if any item fails → revise prompt, regenerate
    if all pass       → emit self_check score block

# Step 7 — Iterative Refinement
on user satisfied → offer style variants / cross-platform adaptation / Carousel expansion / caption (if absent)
on user request   → adjust color / composition / style / format / hook (supports local edit, no full regen)

# Output assembly + dispatch image generation
grid_strategy, caption = finalize_grid_and_caption(format, user_caption, brand_account_context)
                                              # references/grid-and-caption.md
emit design_description (6 items per references/workflow-and-self-check.md §Section 6)
emit caption — separate text block, never rendered into image
emit self_check_score (7-line block)

image = Skill("product-shots-image-gen",
              f"generate: {rendered_prompt} | format={format} | aspect={ratio}")
# Do NOT substitute with direct API call.
assert image.delivered

TOC of Module Files

• references/hard-constraints.md — Section 0 (Dimensions / Safe Zones / Text Overlay / Quality Baseline) + Section 9 (Prompt Banned Words). MUST-level. Loaded at EP Step 0.
• references/industry-dna.md — Section 1: seven industries × six dimensions (Prompt Patch, composition, color, lighting, common errors, social references).
• references/carousel-framework.md — Section 3: 7-Slide Framework + Swipe Psychology + Visual Consistency and Typography (cross-slide anchor template).
• references/workflow-and-self-check.md — Section 2 (Core Workflow seven steps) + Section 4 (Common Mistakes self-audit, 14 items) + Section 6 (Output Format).
• references/grid-and-caption.md — Section 5 (Feed Grid Aesthetics, five strategies) + Section 7 (Caption & Copywriting incl. Social Caption Persona).

Section Index

Goal
Language Rule
0. Hard Constraints                            → references/hard-constraints.md
  Dimensions / Safe Zones / Text Overlay / Quality Baseline
1. Industry Visual DNA                          → references/industry-dna.md
  1.1 Beauty / 1.2 Fashion / 1.3 Tech / 1.4 Lifestyle
  1.5 Travel / 1.6 Food / 1.7 Fitness
2. Core Workflow                                → references/workflow-and-self-check.md §Workflow
  Step 1: Confirm Format and Scenario — Lock Dimensions First
  Step 2: Identify Industry — Auto-Load Visual DNA
  Step 3: Define Engagement Goal
  Step 4: Define the Hook — 3-Second Rule
  Step 5: Composition and Color — Industry × Format Cross-Decision
  Step 6: Self-Check Gate
  Step 7: Iterative Refinement
3. Carousel Standards (Applies when format = Carousel) → references/carousel-framework.md
  A. The 7-Slide Framework
  B. Swipe Psychology
  C. Visual Consistency and Typography
4. Common Mistakes and Self-Audit (Mandatory Gate)    → references/workflow-and-self-check.md §Self-Audit
5. Feed Grid Aesthetics                               → references/grid-and-caption.md §Grid
  Common Grid Strategies / Grid Rules
6. Output Format                                       → references/workflow-and-self-check.md §Output Format
  Design Description (6 Items) / Caption / Self-Check Score
7. Caption & Copywriting                               → references/grid-and-caption.md §Caption
  In-Image Text vs. Caption / Caption Logic / Social Caption Persona
8. Prompt Banned Words — Prevent Artifacts / Watermarks / Screenshot Look
                                                       → references/hard-constraints.md §Section 9
IMPORTANT！Suggestion

Goal

Design social-platform-native visuals (Feed, Story, Reel, Carousel) across Instagram, TikTok, Facebook, Pinterest, RedNote, LinkedIn, and X/Twitter — platform-correct (dimensions / safe zones), industry-appropriate (seven Visual DNA presets), engagement-tuned (SHARE / SAVE / COMMENT), and grid-coherent (Feed harmony). Every output passes a 14-item self-audit and emits a seven-line self-check score.

Language Rule

Always respond in the user's language.
IF user writes in English THEN respond in English
IF user writes in Chinese THEN respond in Chinese
NEVER switch language unprompted.

IMPORTANT — Suggestion Format

IMPORTANT！Suggestion:
Always use <suggestion> format to guide the user.
Do not ask several questions at once.

Example: use <suggestion> format to provide options like
Instagram / TikTok / Facebook / Pinterest / RedNote / LinkedIn / X (Twitter) / Other platform.

This is the front-end rendering instruction — option sets are wrapped in <suggestion> tags so the UI renders them as clickable chips.

Cross-Skill Notes

• Upstream router: product-shots routes here for the organic-social branch when asset_type ∈ {post, carousel, feed, story, reel} AND is_promotion != True. Promotional / ad creative requests route to product-shots-ad-creative instead.
• Industry Visual DNA presets (seven industries) are shared with product-shots and product-shots-ad-creative (cross-skill consistency).
• Sibling skills: product-shots-main-image / product-shots-detail-page (Amazon product/detail page imagery) — can produce source assets that get repurposed for social posts; product-shots-multi-angle (model 9-angle series) — produces fashion lookbook content that can be reframed for social Carousel. product-shots-ad-creative handles paid promotion / ad-objective creative with platform takedown-risk rules.
• Image generation backend: rendered prompts are dispatched to product-shots-image-gen (the product-shots image-gen engine) which abstracts the underlying API (OmniMaaS / OpenAI / Gemini).

Tooling

The skill emits design prompts and self-check output. Image generation is invoked through product-shots-image-gen (the product-shots image-gen engine) using the prompts produced here.