SKILL.md

$2c

Current-request override — if the user names a specific backend in the current message, use it.

Saved preference — if EXTEND.md sets preferred_image_backend to a backend available right now, use it.

Auto-select (when the preference is auto, unset, or the pinned backend isn't available):

**Codex (imagegen)** — first, inspect your available-skills / tool inventory. If a skill named imagegen is listed, you are running inside Codex and MUST use it: invoke via the Skill tool with skill: "imagegen", passing the saved prompt file's content (plus output path and aspect ratio per Codex imagegen's own args). Codex imagegen is the official raster backend in that runtime and outranks any non-native skill (e.g., baoyu-imagine) unless the user has explicitly pinned a different preferred_image_backend.

Other runtime-native tools — if the runtime exposes a different native image tool (e.g., Hermes image_generate), use it the same way.

Otherwise, if exactly one non-native backend is installed (e.g., baoyu-imagine), use it.

Otherwise (multiple non-native backends with no runtime-native tool), ask the user once — batch with any other initial questions.

If none are available, tell the user and ask how to proceed.

⛔ Never substitute SVG, HTML, canvas, or other code-based rendering for raster image generation. Codex imagegen's own description says it should be used "when the output should be a bitmap asset rather than repo-native code or vector." If you cannot resolve a raster backend via step 3, fall through to step 4 and ask the user — do not silently emit SVG, write inline <svg> markup, or produce HTML/CSS art as a substitute. This applies even if the article/section seems "diagram-like": the consumer skill calling this rule has already decided that a raster image is what it needs.

⛔ Never repair rendered text by painting over a generated bitmap. Do not use ImageMagick, Pillow, Canvas, SVG, HTML/CSS, OCR scripts, or any other programmatic overlay to cover, rewrite, erase, stroke, or replace title/subtitle text inside an already generated cover image. If text is wrong or unclear, regenerate from a corrected prompt, switch to a lower-text or no-title variant, or ask the user which imperfect candidate to keep.

Setting preferred_image_backend: ask forces the step-3 prompt every run regardless of available backends. Users change the pinned backend via the ## Changing Preferences section below.

Prompt file requirement (hard): write each image's full, final prompt to a standalone file under prompts/ (naming: NN-{type}-[slug].md) BEFORE invoking any backend. The backend receives the prompt file (or its content); the file is the reproducibility record and lets you switch backends without regenerating prompts.

Concrete tool names (imagegen, image_generate, baoyu-imagine) above are examples — substitute the local equivalents under the same rule.

Confirmation Policy

Default behavior: confirm before generation.

Treat explicit skill invocation, a file path, matched keywords/presets, EXTEND.md defaults, and any documented auto-selection as recommendation inputs only. None of them 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 to do so, for example: --quick, "直接生成", "不用确认", "跳过确认", "按默认出图", or equivalent wording. 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.

If confirmation is skipped explicitly, state the assumed dimensions / aspect / language / backend in the next user-facing update before generating.

Options

Option

Description

--type <name>

hero, conceptual, typography, metaphor, scene, minimal

--palette <name>

warm, elegant, cool, dark, earth, vivid, pastel, mono, retro, duotone, macaron

--rendering <name>

flat-vector, hand-drawn, painterly, digital, pixel, chalk, screen-print

--style <name>

Preset shorthand (see Style Presets)

--text <level>

none, title-only, title-subtitle, text-rich

--mood <level>

subtle, balanced, bold

--font <name>

clean, handwritten, serif, display

--aspect <ratio>

16:9 (default), 2.35:1, 4:3, 3:2, 1:1, 3:4

--lang <code>

Title language (en, zh, ja, etc.)

--no-title

Alias for --text none

--quick

Skip confirmation, use auto-selection

--ref <files...>

Reference images for style/composition guidance

Five Dimensions

Dimension

Values

Default

Type

hero, conceptual, typography, metaphor, scene, minimal

auto

Palette

warm, elegant, cool, dark, earth, vivid, pastel, mono, retro, duotone, macaron

auto

Rendering

flat-vector, hand-drawn, painterly, digital, pixel, chalk, screen-print

auto

Text

none, title-only, title-subtitle, text-rich

title-only

Mood

subtle, balanced, bold

balanced

Font

clean, handwritten, serif, display

clean

Auto-selection rules: references/auto-selection.md

Galleries

Types: hero, conceptual, typography, metaphor, scene, minimal

→ Details: references/types.md

Palettes: warm, elegant, cool, dark, earth, vivid, pastel, mono, retro, duotone, macaron

→ Details: references/palettes/

Renderings: flat-vector, hand-drawn, painterly, digital, pixel, chalk, screen-print

→ Details: references/renderings/

Text Levels: none (pure visual) | title-only (default) | title-subtitle | text-rich (with tags)

→ Details: references/dimensions/text.md

Mood Levels: subtle (low contrast) | balanced (default) | bold (high contrast)

→ Details: references/dimensions/mood.md

Fonts: clean (sans-serif) | handwritten | serif | display (bold decorative)

→ Details: references/dimensions/font.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

Progress Checklist

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

Flow

Input → [Step 0: Preferences] ─┬─ Found → Continue

                               └─ Not found → First-Time Setup ⛔ BLOCKING → Save EXTEND.md → Continue

        ↓

Analyze + Save Refs → [Output Dir] → [Confirm: 6 Dimensions] → Prompt → Generate → Complete

                                              ↓

                                     (skip if --quick or all specified)

Step 0: Load Preferences ⛔ BLOCKING

Check EXTEND.md in priority order — the first one found wins:

Priority

Path

Scope

.baoyu-skills/baoyu-cover-image/EXTEND.md

Project

${XDG_CONFIG_HOME:-$HOME/.config}/baoyu-skills/baoyu-cover-image/EXTEND.md

XDG

$HOME/.baoyu-skills/baoyu-cover-image/EXTEND.md

User home

Result

Action

Found

Load, display summary → Continue

Not found

⛔ Run first-time setup (references/config/first-time-setup.md) → Save → Continue

CRITICAL: If not found, complete setup BEFORE any other steps or questions.

Step 1: Analyze Content

Save reference images (if provided) → references/workflow/reference-images.md

Save source content (if pasted, save to source.md)

Analyze content: topic, tone, keywords, visual metaphors

Deep analyze references ⚠️: Extract specific, concrete elements (see reference-images.md)

Detect language: Compare source, user input, EXTEND.md preference

Determine output directory: Per File Structure rules

⚠️ People in Reference Images:

If reference images contain people who should appear in the cover:

**Model supports --ref** (default): Copy image to refs/, pass via --ref at generation. No description file needed — the model sees the face directly.

**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.

See reference-images.md for full decision table.

Step 2: Confirm Options ⚠️

Hard gate: this step is mandatory per the [Confirmation Policy](#confirmation-policy) — Steps 3–4 cannot start until the user confirms here (or explicitly opts out with --quick / quick_mode: true / equivalent wording in the current request).

**MUST use AskUserQuestion tool** to present options as interactive selection — NOT plain text tables. Present up to 4 questions in a single AskUserQuestion call (Type, Palette, Rendering, Font + Settings). Each question shows the recommended option first with reason, followed by alternatives.

Full confirmation flow and question format: references/workflow/confirm-options.md

Condition

Skipped

Still Asked

--quick or quick_mode: true

6 dimensions

Aspect ratio (unless --aspect)

All 6 + --aspect specified

All

None

Step 3: Create Prompt

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

CRITICAL - References in Frontmatter:

Files saved to refs/ → Add to frontmatter references list

Style extracted verbally (no file) → Omit references, describe in body

Before writing → Verify: test -f refs/ref-NN-{slug}.{ext}

Reference elements in body MUST be detailed, prefixed with "MUST"/"REQUIRED", with integration approach.

Step 4: Generate Image

Backup existing cover.png if regenerating

Select backend via the ## Image Generation Tools rule at the top: use whatever is available; if multiple, ask the user once. Do this once per session before any generation.

Write the full final prompt to prompts/01-cover-[slug].md (hard requirement) BEFORE invoking the backend.

Process references from prompt frontmatter:

direct usage → pass via --ref (use ref-capable backend)

style/palette → extract traits, append to prompt

Generate: Call the chosen backend with the prompt file, output path, aspect ratio

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

Action

Steps

Regenerate

Backup → Update prompt file FIRST → Regenerate

Change dimension

Backup → Confirm new value → Update prompt → Regenerate

Text correction policy:

If the title/subtitle is misspelled, garbled, hard to read, or visually weak, do not patch the bitmap with code.

For text-correction regenerations, write a new prompt file and a new output path so the flawed candidate is preserved for comparison.

Post-processing is limited to crop, resize, compression, or format conversion that does not alter text or the main 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 noted in Step 0. Three ways to change it:

Edit directly — open EXTEND.md and change fields. Full schema: references/config/preferences-schema.md.

Reconfigure interactively — delete EXTEND.md (or ask "reconfigure baoyu-cover-image preferences" / "重新配置"). The next run re-triggers first-time setup.

Common one-line edits:

preferred_image_backend: auto — default; runtime-native tool wins, falls back to the only installed backend, asks only if multiple non-native are present.

preferred_image_backend: codex-imagegen — pin to Codex's built-in.

preferred_image_backend: baoyu-imagine — pin to the baoyu-imagine skill.

preferred_image_backend: ask — confirm backend every run.

watermark.enabled: true, preferred_type, preferred_palette, preferred_rendering, default_aspect, quick_mode: true, language — shift the auto-selection defaults and confirmation flow.