SuperAPI AI Directory
Partner中文AI newsFavorites
Skillslinyuebanzi-knowledge-card

linyuebanzi-knowledge-card

Design & creation

View repository

Updated 11-04 23:11

About this skill

Subject Knowledge Card Generation

Image Generation Method

Directly use the current platform's built-in image generation capability, without calling any external scripts or third-party APIs, and without requiring users to configure an API Key.

  • In Doubao Work: use Doubao Work's built-in image generation capability to produce the image
  • In other Agent platforms: use that platform's built-in image generation tool
  • Call parameters: the complete prompt (produced in Step 4), aspect ratio 9:16, generate 1 image each time
  • If the platform does not support 9:16, use the closest vertical aspect ratio (e.g., 3:4), and tell the user the actual ratio used

When the current platform has no built-in image generation capability, do not call external scripts or third-party APIs; directly tell the user: this Skill needs to be used in a platform with built-in image generation, such as Doubao Work; if using an environment such as Claude Code, please switch to linyuebanzi-edu-infographic(external API image generation version).

Image generation consumes platform quota. For a single image, generate it directly; before batch generation, first tell the user how many images will be generated in total (see Step 9).

Workflow

Step 1: Parse user input

User input examples:

  • "声现象" → Automatically inferred: Grade 8 Physics · Chapter 2
  • "初二物理 光现象" → Directly extracted: Grade 8 Physics · Chapter 4
  • "帮我做一张关于杠杆原理的知识卡片" → Inferred: Grade 8 Physics · Simple Machines
  • "勾股定理" → Inferred: Grade 8 Mathematics
  • "光合作用" → Inferred: Grade 7 Biology

Extract fields:

  • topic: knowledge point name
  • subject: subject
  • grade: grade (automatically inferred if not provided)
  • edition: textbook edition (defaults to PEP; different editions arrange chapters differently, so if the user specifies an edition, follow the user's)
  • chapter: chapter number (if not provided, infer according to edition ; when uncertain, it is better to omit the chapter number than to print a wrong one on the image)
  • key_concepts: 3-5 core sub-concepts (automatically broken down)
  • focus: key points additionally specified by the user (e.g., "focus on experiments," "add a common mistake"), leave blank if none

Step 2: Break down sub-concepts

  1. First check references/illustration_guide.md; for matched knowledge points, directly use the sub-concepts, illustration descriptions, "must include," and "common mistakes" from the library
  2. If there is no match, break out 3-5 sub-concepts according to the textbook, andfill in on the spotthe "must include" and "common mistakes" for each sub-concept (refer to references/accuracy_checklist.md for derivation by the corresponding subject)
  3. For each sub-concept, prepare: title, 1-2 sentence explanation, illustration description, formula or key relationship (if any)

Step 3: Produce an outline and wait for user confirmation

Before generating images, first send the outline to the user,and generate images only after the user confirms. Teachers know best what they want to teach, and this step also avoids wasting quota on incorrect images.

Outline format:

【声现象】初二物理 · 第二章（人教版）· 色卡：陶土橙

1. 声音的产生：声音由物体振动产生，振动停止，发声停止
   插图：三个音叉，中间一个带振动箭头，右侧声波弧线
2. 声音的传播：传播需要介质，真空不能传声
   插图：敲鼓，同心圆声波扩散
3. ……

底部：学习提示框「音调看疏密，响度看高低」

确认就回复「出图」；要改直接说，比如「第 3 点加一个回声的例子」。
  • User replies with revision suggestions → revise the outline and confirm once more
  • User says "just generate directly" or "no need to confirm" at the start → skip this step
  • For batch generation, list the outlines for all chapters at once, and the user confirms them in one go

Step 4: Determine the theme color and build the prompt

Theme color: Use low-saturation (Morandi) tones throughout—do not give the image generation model coarse-grained color names like "blue family" (it will default to outputting high-saturation "AI infographic blue"), instead givespecific color palettes: primary color + background color + neutral color, each with a hex anchor. For different chapters of the same subject, rotate the primary color among the optional palettes, with no repeats within the same batch.

Palette Primary (accent) Background (bg tint) Neutral Applicable to
Terracotta Orange terracotta #C97E4E cream #F7F1E6 slate blue-grey #3A506B Physics (sound/heat), Chemistry (organic)
Deep Gray-Blue slate blue #3A506B misty blue-white #EDF3F8 warm grey #8A7D6B Physics (electricity/mechanics), English
Soft Cyan-Green muted teal #4A9E9B pale green-white #EAF1E3 grey olive #8E9678 Physics (light/lenses), Chemistry
Misty Purple misty purple #7A6296 ivory #F7F3ED grey blue #6B7889 Mathematics
Deep Tea Green deep tea green #5F6B4F pale green-white #EAF1E3 light khaki #D9C9A3 Biology
Caramel Brown caramel brown #8A624A cream beige #F4E6C8 grey coffee #6F5F4A Chinese, History
Sage Green sage green #8FA08A oat #F2EAD9 grey olive #8E9678 Biology (plants), English

Color rules:

  • The primary color is used only ontitles, number badges, keywords, and formula highlights(about 10% of the image); card backgrounds and the background use light variants of the background color
  • The neutral color is for secondary text, structural lines, and explanatory annotations
  • Prohibit high-saturation candy colors, neon colors, and pure #FF hues

Build the prompt: Read references/prompt_template.md to get the prompt skeleton, and fill in the following variables:

  • {topic} → knowledge point name
  • {grade} → grade
  • {subject} → subject
  • {chapter} → chapter (if uncertain, omit the entire segment)
  • {theme_color} → primary color (English name + hex, e.g. misty purple #7A6296）
  • {bg_tint} → background color (English name + hex)
  • {neutral_color} → neutral color (English name + hex)
  • {num_sections} → number of sub-concepts
  • {sections_content} → the titles, illustration descriptions, and explanatory text of each sub-concept confirmed in Step 3; write the "common mistakes" column as prohibitions in the prompt

Step 5: Subject accuracy self-check (prompt layer)

Before calling the image generation, check the built prompt. This is the first line of defense, blocking "errors written in the prompt itself."

Data sources: Read the following three files at the same time and cross-check them against the current prompt:

  • references/accuracy_checklist.md: General checklist + subject-specific checklists
  • references/illustration_guide.md: The "must include" and "common mistakes" columns for each knowledge point, checked item by item
  • references/errata.md: Historical errata records, to confirm no repeated pitfalls

Check order:

  1. General check(all subjects): typos, formula variable labeling, grade matching (out-of-syllabus check), chapter number consistent with textbook version, numerical reasonableness
  2. Subject-specific check: Go through the checklist for the corresponding subject in accuracy_checklist.md item by item. For physics, check controlled-variable conditions/circuit conventions/scalars and vectors; for mathematics, check theorem applicability conditions/graphic labeling consistency; for chemistry, check balancing/reaction conditions/↑↓ symbols; for biology, check the four elements of processes/arrow directions; for Chinese, check the original text of famous quotes/correspondence of authors to dynasties; for history, check dates/correspondence of people and events; for English, check spelling/part of speech/example sentence grammar

If problems are found, correct the prompt first before calling the image generation.

Step 6: Generate the image using the platform's built-in image generation capability

Call the current platform's built-in image generation according to the "Image generation method" above:

  • Prompt: the complete prompt after Step 4 construction and Step 5 self-check,pass it in as-is; do not trim it yourself
  • Aspect ratio: 9:16 vertical
  • Quantity: 1 image
  • File name:{topic}_信息图; for batches add a sequence number, e.g. 01_机械运动_信息图

After generation, save the image and show it to the user; if the user specified a save location (such as a folder), save it there.

Step 7: Visual review of the finished image (image level)

This is the second line of defense, guarding against "the prompt is right but the image is wrong"—image generation models may render garbled text, typos, incorrect formula sub/superscripts, reversed arrows, or mixed-up circuits.Every image must be reviewed after generation; no delivery without review.

  1. View the image just generated (look at it directly, or open the image file with an image-reading tool)
  2. Against the references/accuracy_checklist.md "key points for visual review of the finished image" at the bottom, check item by item what isactually renderedin the image:
    • All Chinese text: no garbled characters, no typos, no pseudo-Chinese characters with missing strokes, no repeated characters
    • Formulas/equations/numbers: exactly consistent with the prompt, sub/superscripts correct, units not lost
    • Diagram logic: arrow directions, label positions, and structural relationships consistent with the prompt description
    • Then go through the image content again according to that subject's subject-specific checklist
  3. Judgment and handling:
    • Pass → deliver
    • Fail → add emphasis instructions to the prompt targeting the error points (e.g. The formula must read exactly "a² + b² = c²" with correct superscripts), and regenerate. Retry at most 2 times
    • If still failing after 2 retries → deliver the best one among them, butyou must clearly tell the user where the image is wrong, and never silently deliver a wrong image to students
  4. When delivering, attach a sentence with the review result, e.g. "Checked: text, formulas, and diagrams are all correct" or "The arrow direction in block 2 is wrong; manual correction is recommended"

Step 8: One-sentence user modification

After delivery, the user may say "change the illustration in block 3 to xx", "change the title to xx", or "this is wrong here, it should be xx":

  1. Only change the corresponding part of the prompt; keep the rest unchanged
  2. The modified content must likewise go through the Step 5 self-check, then be regenerated, and go through the Step 7 review
  3. Errata feedback: knowledge errors pointed out by the user, or knowledge errors found during review (excluding pure rendering garbling), should be appended in the format to references/errata.md(when the current environment allows writing to the Skill directory), so the next generation automatically avoids the pitfall; when writing is not possible, list the errata entries in the reply so the user can add them themselves

Step 9: Batch generation

When the user requests multiple knowledge points (e.g. "all chapters of the first semester of eighth-grade physics"):

  1. Break it down into a list of independent knowledge points and tell the user how many images will be produced in total
  2. Follow Step 3 and list the outlines of all chapters at once for the user to confirm
  3. Produce the first one as a sample card, and after the user confirms they are satisfied with the style, continue with the rest; if the user says "just produce them all directly", go ahead without stopping
  4. Assign a different theme color to each image (no repetition within the same batch), with a unified style but varied colors
  5. Each image independently goes through the Step 5 self-check and Step 7 visual review, without skipping due to batching
  6. Final summary: report the review result for each image (passed / passed after retry / delivered with flaws and the specific issues)

Usage example

# 单个知识点
声现象，做张信息图
帮我做一张物态变化的知识卡片

# 指定学科、年级、教材版本
初二物理 光现象（沪科版）

# 跨学科
勾股定理
光合作用

# 批量
帮我把初二物理上学期的知识点全部做成信息图
初二物理 声现象、物态变化、光现象、透镜，分别做一张

Notes

  • All text must be in Chinese (except example sentences and words for English subjects)
  • Illustrations should mainly be simple diagrams; do not use real photographs
  • Limit each infographic to 3-5 sub-concepts
  • Formulas must clearly label the meaning of variables
  • Choose one of three for the bottom summary area: key term badges / study tip box / real-life example strip
  • Multiple images generated in the same batch must use different color schemes
  • The audience is students; accuracy takes priority over aesthetics: Each image must complete the Step 5 prompt self-check + Step 7 visual review of the finished image before delivery; if an unfixable factual error is found, the user must be informed and it must not be delivered silently
  • AI handles layout and illustrations; whether the knowledge points are correct is ultimately subject to the teacher and textbook; when delivering, remind the user to check it once before sending it to students
  • Chapter numbers are labeled according to the PEP edition by default; if the user uses textbooks from other editions, please specify in the input
  • This Skill only generates static infographics