baoyu-cover-image

Cover Image Generator

User Input Tools

When this skill prompts the user, follow this tool-selection rule (priority order):

1. Prefer built-in user-input tools exposed by the current agent runtime — e.g., AskUserQuestion, request_user_input, clarify, ask_user, or any equivalent.
2. Fallback: if no such tool exists, emit a numbered plain-text message and ask the user to reply with the chosen number/answer for each question.
3. Batching: if the tool supports multiple questions per call, combine all applicable questions into a single call; if only single-question, ask them one at a time in priority order.

Concrete AskUserQuestion references below are examples — substitute the local equivalent in other runtimes.

Image Generation Tools

Resolve the image backend in this order:

1. User names a backend in the current message → use it.
2. `preferred_image_backend` in EXTEND.md names a backend available now → use it. ask means ask every run.
3. Auto-select (preference auto, unset, or the pinned backend is unavailable), first match wins:

• Codex `imagegen` skill is in your available skills → you are in Codex and MUST use it: Skill with skill: "imagegen", passing the prompt file's content plus output path and aspect ratio. It outranks non-native skills such as baoyu-image-gen.
• Codex via `codex exec` (`codex-imagegen`): no native imagegen, but the codex CLI is on PATH with an active codex login → baoyu-image-gen --provider codex-cli (preferred), or the bundled wrapper if baoyu-image-gen is unavailable. Load references/codex-imagegen.md only for this branch.
• Cursor `GenerateImage` → use it (outranks non-native skills). It has no aspect-ratio parameter, so state the aspect ratio / dimensions in the description prompt; it saves to a tool-managed location, so copy the file to the expected output path (e.g., outputs/.../NN-xxx.png). Reference images go in reference_image_paths.
• Other runtime-native image tool (e.g., Hermes image_generate) → use it the same way.
• Exactly one non-native backend installed (e.g., baoyu-image-gen) → use it.
• Several non-native backends → ask the user once, batched with other initial questions.

1. None available → tell the user and ask how to proceed.

⛔ Raster only: never substitute SVG, HTML/CSS, canvas, or other code-based rendering for a raster image, even for diagram-like content. If no raster backend resolves, go to step 4.

⛔ No bitmap text patching: never use ImageMagick, Pillow, Canvas, SVG, HTML/CSS, OCR scripts, or any other overlay to cover, rewrite, erase, stroke, or replace title/subtitle text inside an already generated cover image. Regenerate from a corrected prompt, switch to a lower-text or no-title variant, or ask the user which imperfect candidate to keep.

Prompt file (hard): before invoking any backend, write each image's full, final prompt to its own file under prompts/ (NN-{type}-[slug].md) and pass that file (or its content) to the backend. It is the reproducibility record and lets you switch backends without rewriting prompts.

Change the pinned backend via ## Changing Preferences below. Tool names above are examples; use the local equivalent in other runtimes.

Confirmation Policy

Default: confirm before generation.

• Explicit skill invocation, a file path, matched keywords/presets, EXTEND.md defaults, and any documented auto-selection are recommendation inputs only — none authorizes skipping confirmation.
• Do not start Step 3 or Step 4 until the user confirms the dimensions / aspect / language / backend choices.
• Skip confirmation only when the current request explicitly says so: --quick, "直接生成", "不用确认", "跳过确认", "按默认出图", or equivalent. quick_mode: true in EXTEND.md counts as a standing explicit opt-out — set it only when you want every run to skip Step 2.
• When skipped, state the assumed dimensions / aspect / language / backend in the next user-facing update before generating.

Options

Five Dimensions

Values are listed in Options. Auto-selection: references/auto-selection.md. Compatibility matrices: references/compatibility.md. Icon vocabulary: references/visual-elements.md.

File Structure

Output directory per default_output_dir preference:

• same-dir: {article-dir}/
• imgs-subdir: {article-dir}/imgs/
• independent (default): cover-image/{topic-slug}/

<output-dir>/
├── source-{slug}.{ext}    # Source files
├── refs/                  # Reference images (if provided)
│   ├── ref-01-{slug}.{ext}
│   └── ref-01-{slug}.md   # Description file
├── prompts/cover.md       # Generation prompt
└── cover.png              # Output image

Slug: 2-4 words, kebab-case. Conflict: append -YYYYMMDD-HHMMSS

Workflow

Cover Image Progress:
- [ ] Step 0: Check preferences (EXTEND.md) ⛔ BLOCKING
- [ ] Step 1: Analyze content + save refs + determine output dir
- [ ] Step 2: Confirm options (6 dimensions) ⚠️ unless --quick
- [ ] Step 3: Create prompt
- [ ] Step 4: Generate image
- [ ] Step 5: Completion report

Step 0: Load Preferences ⛔ BLOCKING

First match wins:

Found → load, display summary. Not found → ⛔ run first-time-setup and save, BEFORE any other step or question.

Step 1: Analyze Content

1. Save reference images (if provided) → references/workflow/reference-images.md
2. Save source content (if pasted, save to source.md)
3. Analyze content: topic, tone, keywords, visual metaphors
4. Deep analyze references ⚠️: extract specific, concrete elements (see reference-images.md)
5. Detect language: compare source, user input, EXTEND.md preference
6. Determine output directory per File Structure

People in reference images who should appear in the cover:

• Model supports `--ref` (default): copy image to refs/, pass via --ref at generation. No description file needed.
• Model does NOT support `--ref` (Jimeng, Seedream 3.0): create refs/ref-NN-{slug}.md with per-character description (hair, glasses, skin tone, clothing); embed as MUST/REQUIRED instructions in prompt text.

Step 2: Confirm Options ⚠️

Hard gate per the Confirmation Policy (opt-out: --quick / quick_mode: true / equivalent wording in the current request).

MUST use `AskUserQuestion` (interactive selection, not plain-text tables): up to 4 questions in one call (Type, Palette, Rendering, Font + Settings), each with the recommended option first plus reason. --quick / quick_mode still asks aspect ratio unless --aspect is given; all 6 dimensions + --aspect specified skips everything. Full flow: references/workflow/confirm-options.md.

Step 3: Create Prompt

Save to prompts/cover.md per references/workflow/prompt-template.md.

• Frontmatter references: only files saved to refs/ (verify test -f refs/ref-NN-{slug}.{ext}); verbally extracted style → omit references, describe in body.
• Reference elements in the body MUST be detailed, prefixed with "MUST"/"REQUIRED", with an integration approach.

Step 4: Generate Image

1. Backup existing cover.png if regenerating
2. Select backend once per session via ## Image Generation Tools.
3. Write the full final prompt to prompts/01-cover-[slug].md (hard requirement) BEFORE invoking the backend.
4. Process references from prompt frontmatter: direct → pass via --ref (ref-capable backend); style/palette → extract traits, append to prompt.
5. Generate with the prompt file, output path, aspect ratio. For codex-imagegen, see references/codex-imagegen.md.
6. On failure: auto-retry once

Step 5: Completion Report

Cover Generated!

Topic: [topic]
Type: [type] | Palette: [palette] | Rendering: [rendering]
Text: [text] | Mood: [mood] | Font: [font] | Aspect: [ratio]
Title: [title or "visual only"]
Language: [lang] | Watermark: [enabled/disabled]
References: [N images or "extracted style" or "none"]
Location: [directory path]

Files:
✓ source-{slug}.{ext}
✓ prompts/cover.md
✓ cover.png

Image Modification

Text correction: never patch a misspelled, garbled, or weak title/subtitle in the bitmap with code. Write a new prompt file and a new output path so the flawed candidate is kept for comparison. Post-processing is limited to crop, resize, compression, or format conversion that doesn't alter text or composition.

Composition Principles

• Whitespace: 40-60% breathing room
• Visual anchor: Main element centered or offset left
• Characters: Simplified silhouettes; NO realistic humans
• Title: Use exact title from user/source; never invent

Changing Preferences

EXTEND.md lives at the path from Step 0. Edit it directly (schema: references/config/preferences-schema.md; watermark: references/config/watermark-guide.md), or delete it / ask "reconfigure baoyu-cover-image preferences" / "重新配置" to re-run first-time setup. Common edits:

• preferred_image_backend: auto (default; runtime-native wins, else the only installed backend, ask if multiple non-native) | codex-imagegen | baoyu-image-gen | ask (confirm every run)
• watermark.enabled: true, preferred_type, preferred_palette, preferred_rendering, default_aspect, quick_mode: true, language — shift auto-selection defaults and the confirmation flow.