BruceL017/post-illustration-images

Post Illustration Images

Core Rule

Do not reduce this workflow to "analyze content -> choose template -> generate image". Stability depends on the middle layers:

content analysis
-> content type and expression need
-> suite-level Style Spec
-> anchor selection
-> shot list
-> per-image content structure
-> per-image visual metaphor
-> selected visual style constraints
-> brand slot reservation when Brand Plugin is enabled
-> single-image prompt without model-drawn brand or page badge
-> native Codex image generation
-> optional Brand Plugin overlay using the selected Style Spec slot
-> QA and fallback
-> saved assets and delivery notes

Always read the content before choosing a visual style. Use one suite-level style for the whole image set. Vary each image only through its content structure and visual metaphor.

Architecture Boundary

Style Spec is the authority for each platform visual template. It controls canvas size, aspect ratio, fixed color values, layout, safe areas, negative constraints, and any fixed component slots such as a brand slot.

Brand Plugin is optional and pluggable. It only decides whether a brand is enabled and which brand asset to use. It must not decide canvas size, color palette, coordinates, or platform layout. When enabled, the Brand Plugin must obey the selected Style Spec's brand slot. If the selected style has no brand slot, do not improvise a global placement.

The image model must not draw logos, TF, Tranfu, watermarks, page-number badges, or other fixed brand components. Those are either omitted or added after generation by a deterministic overlay step.

Supported Scope

Strong support:

• WeChat official account body illustrations.
• Xiaohongshu note illustrations, including cover/content carousel images.
• Zhihu post illustrations.
• Method, workflow, concept, comparison, checklist, boundary, and decision-flow explainers.

Conditional support:

• Image sets requiring real product screenshots or factual data: ask for the material or state the assumption.
• Explicit style names: use the matching style file if it exists.
• Explicit image count: override the selected style's default count.

Out of scope:

• Pure photography, portrait retouching, product mockups, or photorealistic brand campaigns.
• Tasks where exact long text inside images is mandatory. Native image generation may distort text; use short labels and retry with less text if needed.
• Any request to route through other image-generation skills unless the user explicitly asks for them.

Required References

Use references by need, not all at once:

• Start with references/style-index.md to identify platform and candidate style files.
• Read the selected file under references/styles/ completely before compiling prompts.
• Read references/brand.md unless the user explicitly disables the Brand Plugin.
• Read references/content-structures.md when selecting per-image expression structures.
• Read references/prompt-compiler.md before writing final generation prompts.
• Read references/qa-checklist.md before delivery or regeneration.
• Read references/product-design.md only when maintaining or extending this skill.

Workflow

1. Intake

Determine the platform, source content, requested output, explicit style, explicit count, output location, and whether the Brand Plugin is disabled.

If platform cannot be inferred, ask one concise question. If platform is clear and content is available, proceed without asking for confirmation.

Exit when:

• Platform is known.
• Source content is available.
• Any explicit count/style/brand override is captured.

2. Content Analysis

Read the full source content. Summarize the main line in 3-5 sentences, then classify the expression need:

• Concept explanation
• Workflow/process
• Before-after comparison
• Layered framework
• Common mistakes or boundaries
• Decision tree
• Checklist or summary
• Story/state transition

Exit when the core claim, audience value, and visualizable parts are clear.

3. Select Suite-Level Style Spec

Use references/style-index.md to choose one style for the whole set.

Rules:

• Explicit user style wins if the style exists.
• Otherwise choose the best platform style based on content type and expression need.
• Do not decide image count, aspect ratio, palette, safe area, or fixed component slots before selecting the style. These come from the style file or its machine-readable Style Spec unless the user explicitly overrides them.
• For Zhihu, use the fixed Zhihu style unless future style files add alternatives.

Exit when one style file is selected and read completely.

4. Select Anchors

Choose only the content points that deserve images. Do not create images to fill a quota.

Good anchors include:

• Core judgment
• Cognitive turn
• Input-output loop
• Before-after contrast
• Common mistake
• Process breakpoint
• Role/state change
• Method layer
• Trust, conversion, or handoff relationship

Exit when each selected anchor can answer: "why does this need an image?"

5. Build Shot List

The shot list is mandatory. Each image must contain:

• Placement or sequence
• Topic
• One core meaning
• Content expression structure
• Visual metaphor
• Main actor/object action
• Suggested elements
• Short labels
• Fixed component reservations from the selected Style Spec, if any
• QA risk

Each image must express exactly one core idea. If one image contains multiple ideas, split or delete.

6. Compile Prompts

For each image, combine:

• User content anchor
• Selected suite-level style file
• Per-image content structure
• Per-image visual metaphor
• Style Spec safe areas and fixed component reservations
• Brand Plugin status from references/brand.md
• Native Codex generation constraints
• Negative constraints from the style and QA references

Generate one prompt per image. Do not ask the image model to create an entire multi-image set in one call.

If a style file contains batch language such as "generate the whole set", treat it as planning guidance for count, sequence, and consistency. Still compile and generate one image at a time.

The prompt must explicitly forbid model-drawn logos, TF, Tranfu, watermarks, and page-number badges. When the Brand Plugin is enabled, ask the model to keep the selected Style Spec's brand slot clear so the real brand asset can be overlaid after generation.

7. Generate One Image At A Time

Use native Codex image generation only. Do not invoke other image skills by default.

After each generation:

• Save the image into the current project, not into the skill folder.
• Recommended output path: post-illustration-output/<content-slug>/.
• Save native generated images under images/unbranded/ when Brand Plugin is enabled.
• Save post-overlay deliverables under images/branded/ when Brand Plugin is enabled.
• If Brand Plugin is disabled, save deliverables under images/.
• Use ordered filenames, for example 01-cover.png, 02-process-breakpoint.png.

7.5 Apply Brand Plugin Overlay

If the Brand Plugin is enabled and the selected Style Spec defines a brand slot:

• Use the real brand asset from references/brand.md.
• Overlay the asset after image generation. Do not ask the image model to redraw the logo.
• Use the selected Style Spec's slot, size, and safe-area rules.
• Keep the unbranded generated image unless the user explicitly asks to replace it.
• For the Xiaohongshu journal style, use scripts/apply-brand-overlay.mjs with references/styles/xhs-style-journal.spec.json.
• The overlay script has no npm dependency, but it requires rsvg-convert to be available on the machine.

8. QA And Fallback

Check every image against references/qa-checklist.md.

If an image fails, identify the reason before retrying:

• Text error: reduce text, keep only short labels, regenerate.
• Too many ideas: split or remove anchors.
• Style drift: strengthen selected style source and negative constraints.
• Weak metaphor: rewrite physical action and object.
• Brand was drawn by the model: regenerate with stronger "no logo/no TF/no Tranfu" constraints, then apply the Brand Plugin overlay.
• Brand overlay blocks content: use the Style Spec's reserved brand slot; if content occupies that slot, regenerate with a clearer reserved area.
• Layout too empty or crowded: adjust structure, not the whole style.

9. Save And Deliver

Create or update manifest.md in the output folder with:

• File
• Platform
• Style
• Brand Plugin enabled/disabled
• Style QA status
• Brand Plugin QA status, if enabled
• Placement/sequence
• Core meaning
• Structure
• Visual metaphor
• QA status

Final response must include:

• Output path.
• Number of images.
• Which style was used.
• Each image's usage.
• QA conclusion and any residual risk.

Non-Negotiables

• Analyze content before style selection.
• Use one suite-level visual style for the whole set.
• Never skip the shot list.
• Every image expresses one core meaning.
• Content expression structure organizes information.
• Visual metaphor turns abstraction into a concrete scene.
• Suite-level Style Spec controls platform appearance, dimensions, colors, layout, safe areas, and fixed component slots.
• Brand Plugin is enabled by default only when the selected Style Spec defines a brand slot; it must obey that slot and must not behave as a global visual system.
• The image model must not draw brand logos or page-number badges.
• Generate and QA one image at a time.
• Save generated assets outside the skill folder.