Organize content such as screenshots, board notes, courseware, flowcharts, and structure diagrams into text notes suitable for review and secondary copying.
Purpose
Extract text information from images.
Reconstruct the structural relationships beside the image.
Organize scattered visual information into plain text output suitable for note-taking.
By default, output an ASCII block diagram plus textual explanation, ensuring it can be copied directly into documents, memos, Markdown, or chat records. Unless the user explicitly requests plain text, do not omit the block diagram.
When this skill is useful
Prioritize handling the following types of requests.
Convert images into text.
Organize the diagram and text together into notes.
Rewrite the structure diagrams in the courseware as plain text.
Organize addressing modes, flowcharts, hierarchy diagrams, and relationship diagrams into copyable block diagrams.
Output in the order of title → block diagram → explanation.
Convert the content into exam review drafts, recitation drafts, or class note drafts.
Even if the user only says to organize into notes, as long as there are obvious structural relationships in the image, still return a block diagram + textual explanation by default.
Core workflow
First read the image completely.
First identify the text that can be reliably recognized, then supplement the structure; do not start by guessing details.
If the image has explanatory text on the left and a structure diagram on the right, extract them separately, then merge them into a single unified note.
Prioritize preserving the terminology in the original image.
When encountering blurry areas, output only high-confidence content. For characters that cannot be seen clearly, directly state that they are blurry; do not fill in professional terms by imagination.
If the user wants a note version, by default provide title + ASCII block diagram + explanation; only omit the block diagram when the user explicitly says they only want plain text.
If the user wants a block diagram version, prioritize providing the ASCII block diagram.
If the user has corrected the format, strictly follow that order in subsequent output.
Output modes
Choose the closest output based on the user's wording.
Plain text notes
Suitable for requests such as convert images to text, organize into text, or make into notes.
The output goal is to combine the text and graphical meaning in the image into readable text.
The suggested structure is as follows.
标题
核心内容整理
含义
特点
补充说明
Block diagram notes
Suitable for requests such as return a block diagram, draw it as a text block diagram, or redraw the structure diagram in text.
Use the following order by default.
First write the title.
After a line break, provide a single block diagram.
Then after another line break, write the explanation.
Do not explain first and then provide the diagram.
Do not pile multiple knowledge points into one large page unless the user explicitly requests an overview version. By default, one knowledge point corresponds to one separate block diagram; do not merge multiple addressing modes into one summary table.
The recommended template is as follows.
寄存器间接寻址
┌────┬────┐
│ OP │ R │
└────┴────┘
│
▼
┌────────┐
│ 寄存器 │ ─────→ 操作数地址
└────────┘ │
▼
┌────────┐
│ 主存 │ ───→ 操作数
└────────┘
含义是……。特点是……。
Review draft
Suitable for requests such as help me organize it into a recitation version, exam note version, or ultra-simplified version.
Keep only the minimum necessary information.
Write one sentence per meaning clearly.
Avoid piling up details.
Diagram reconstruction rules
When translating visual relationships into textual relationships, prioritize preserving these elements.
Boxes.
Arrow directions.
Hierarchical relationships.
Sequential access order.
The difference between intermediate quantities and final quantities.
For example, this kind of chain must be clearly layered.
Instruction → address code → main memory → operand
Instruction → address code → main memory → operand address → main memory → operand
Instruction → register number → register → operand
Instruction → register number → register → operand address → main memory → operand
A block diagram does not need to be pixel-for-pixel identical to the original image.
The key points are correct relationships, correct order, correct terminology, and ease of copying.
OCR judgment rules
Directly organize clear content.
For blurry titles, only carefully fill them in when the context is strong enough.
For characters that cannot be confirmed, do not pretend you saw them clearly.
If only a few individual words are blurry but the overall structure is very clear, you can note in the explanation that a certain word is blurry, and the structural relationships have been organized based on the clear parts.
If the user later corrects a certain term or format, take the user's version as authoritative and immediately follow it.
Style rules
Default to concise.
Do not output the OCR process.
Do not repeat the same information.
Do not add knowledge extensions that are not in the image just to appear complete.
When the user does not ask for it, do not additionally explain background textbook knowledge.
What the user wants is copyable notes, not an image analysis report.
Response patterns
Pattern A: The user only wants text
Directly return the organized plain-text notes, without drawing a block diagram.
Pattern B: The user says to organize into notes, but does not restrict the format
By default, return title + one or more necessary block diagrams + explanation.
If one image contains multiple knowledge points, each knowledge point must be split into a separate subsection, each with its own title, block diagram, and explanation; do not first draw an overview tree and then expand it.
If the image has explanatory text on the left and a structural diagram on the right at the same time, merge them into unified notes.
Pattern C: The user wants both diagrams and text converted to text
By default, understand this as wanting copyable notes, and return block diagrams + text explanations.
First give the terminology title.
Then give the corresponding block diagram or structural chain.
Then give the meaning or characteristics.
Pattern D: The user wants a single block diagram
Output only one title corresponding to one block diagram.
Follow the block diagram immediately with an explanatory paragraph.
Do not casually expand other knowledge points of the same kind as well, unless the user explicitly asks for the complete set.
Pattern E: The user gives a format correction
Treat the correction as the final template.
Strictly output according to that template from then on.
For example, when the user requires title first, then block diagram, then introduction, use this order fixedly from then on.
Example transformation
Example 1
Input
Convert the image to text
Output
立即寻址方式
┌────┬────────┐
│ OP │ 操作数 │
└────┴────────┘
特点：操作数直接在指令中，速度快，灵活性差。
Only when the user explicitly says only plain text may the block diagram be removed.
Example 2
Input
Convert the image and the diagram next to it into text, so it is convenient for me to take notes
Output
寄存器间接寻址方式
IR -> OP | R -> 操作数地址 -> 主存 -> 操作数
含义：寄存器中存放的是操作数地址，再按地址到主存取操作数。
Example 3
Input
Using register indirect addressing as an example, do it again
Output
寄存器间接寻址
┌────┬────┐
│ OP │ R │
└────┴────┘
│
▼
┌────────┐
│ 寄存器 │ ─────→ 操作数地址
└────────┘ │
▼
┌────────┐
│ 主存 │ ───→ 操作数
└────────┘
含义是指令中给出寄存器编号，寄存器中保存操作数地址，再按该地址到主存取出操作数。
Final check
Check whether the title appears before the block diagram.
Check whether the block diagram contains only the one knowledge point the current user wants.
Check whether the explanation is placed after the block diagram.
Check whether the terminology is consistent with the image.
Check whether the output can be directly copied into note-taking software.