coffe-d/shenlun.skill
WritingThis page was translated automatically. Technical details may differ from the original. View the original
An in-depth analysis of the essay teaching philosophies of Xiao Ma, Bailu, and Yuan Dong, integrating a practical set of essay question preparation and intelligent grading skills
Overview
Integrating the methodologies of Banyue Tan Bailu, Bilibili's Xiao Ma Ge (short questions), and Yuan Dong (Da Zu Wen), the system rigorously evaluates answers from the perspectives of Shenlun examiners and graders, providing sentence-by-sentence scoring and improvement suggestions for sub-items. Supports**: Test paper photos, question screenshots, and handwritten answer photos can be sent directly, native model reading without the need for pre-transcribed text or external image reading tools. - : Each correction should assign the lost points to the user's local error bank according to ** / - : Record loss points for cross-question comparison like "B1-No Unified Sentence Pattern," not "a question missing a certain scoring point"—only types can accumulate counts and reveal recurrence patterns - : 🟡 Initial appearance (1 time) → 🟠 Recurrence (2 times) → 🔴 Stubborn (≥3 times) - : When the same loss point type appears again, ; Stubborn items are given ** - : Ask "What is my old problem?" "Have I improved recently?", sort and output by stubborn → recurrence → initial appearance, and clearly indicate the improved items - classification system (8 dimensions · fine-grained codes) see references/error-bank-rules.md; error bank/gitignore, - : Test paper photos, question screenshots, scanned material copies, handwritten answer photos can all be sent directly, no need to transcribe into text or connect to external image reading tools - Image reading standards: Inventory of image types → Structured extraction (question number/score/word count/question stem/materials) → Marked on blurry areas for self-check → Transcription confirmation → Downgrade only when unreadable - Proactively checks with users during reading; if not accurate, mark [image blurry] without guessing - Supports multiple images (multiple pages of test paper spliced in page order), mixed images (image in case of conflicts), distinguishing print and handwriting - Compatible with DSH / Claude Code / Codex / Cursor, as long as the selected model supports image input - After submitting content, the exam year, paper type, question number, and original question are automatically located - After successful matching, the actual grading point is used as the scoring standard, skipping general inference - After each correction, the exam bank is automatically stored and gradually expanded (references/real-exams/) - (Including summary methods/problems/causes/experiences/results/changes/characteristics): No summary sentences, key points listed in points, direct copying of original words from the material - (Four subcategories: word and sentence comprehension/viewpoint analysis/phenomenon analysis/relationship analysis): summary sentence key question + dimensional expansion + kickback question stem - : format score + content filtering (selecting key points by essay type) + matching language style - dual-perspective reference answers: Xiao Ma Ge version (original word of material) + Egret version (synonym extraction, light linkage) - default concise format output, detailed scoring as needed - Yuan Dong's "Three Types" Topic Judgment (Single Theme/Dual Theme, AB Type/Dual Theme, ABC Type/Multi-Theme) -...
Overview (machine translated)
Three famous teachers, three mirrors—reflecting the many possibilities of the answer, and also your own kind
The sub-topic references Bai Lu and Xiao Ma Ge, and the major essay follows Yuan Dong—based on current reputation evaluations, not theoretical deductions.
Combining the methodology of three renowned teachers with evidence-based methods, rigorously evaluating answers from the perspectives of essay question setters and graders. Supports dual input of text and images: test paper photos, question screenshots, and handwritten answer photos can be sent directly, with native model reading.
**Advice is given, not judgment. ** Different solutions for the same question are presented for comparison, helping you form a problem-solving methodology that suits you best through comparison.
Quick Start · Sample Effects · Quick Check for Question Types · Architecture and Installation · Functional Modules · Sources and Statements
Get Started Quickly
When to use it? Just say it directly
| You say | Trigger |
|---|---|
| “Help me grade this countermeasure question” | Question correction (dual-perspective answers + grader scoring) |
| “This comprehensive analysis question is each from the perspectives of Little Ma Ge and Bailu.” | Dual-perspective reference answers + difference comparison |
| “This essay helps me check if there’s any off-topic” | Essay Assessment (Theme Type → Four Types of Essays Scoring) |
| “How is the essay I wrote?” (No prompt provided) | First, request the original topic; if you can’t get it, switch to general grading |
| “I just want to change the language, don’t pick typos” | Suggestions for replacement according to the upgraded rules |
| “What are my old habits?” “Have I made progress recently?” | Reading the wrong question bank to diagnose weaknesses |
| “I am preparing for the 2026 National Exam at the prefecture level” | Question Prediction |
**No need to memorize any commands. ** Send questions, answer questions, or send pictures—all are fine—the router will automatically assign the corresponding skills.
Example of the effect
⚠️ > The following is a schematic example (used to illustrate output format), not a real grading record.
Actual grading will be given based on your specific question and answer, with content changing according to the question.
Composition of output from a single correction (Question as an example; Major essay isomorphism, switch to Yuan Dong’s four-category essay scoring):
| # | Output | Description |
|---|---|---|
| ① | Real Exam Recognition Results | Test Paper/Question Number/Question Type/Score/Original Question |
| ②③ | Dual-Perspective Reference Answers | Xiao Ma Ge’s version (fixed structure, integrated elements) + Egret version (structure follows the question stem, element columns are divided) — The two editions have different structures, not a change of wording |
| ④ | Difference Comparison | Which scoring point did each person step on/miss · Differences in classification logic · Applicable scenarios · Individual risks · “Which set suits you best?” |
| ⑤ | Grader Evaluation | Scoring points are checked item by item + total score (using the actual score of the question, not on a 100-point scale) |
| ⑥ | Improvement Suggestions | Five-paragraph format: Points of loss → Expression upgrade → Idioms and famous sayings → Argumentation approach → Optimization demonstration (Only discuss how to change, no diagnosis) |
| ⑦ | Reactivation Reminder | When a certain point type accumulates ≥2, indicate “This is the Nth time” and which question was last time |
Scenario One: Submit questions and correct them
Scenario 2: Submit a Major Essay
Scenario Three: Only Answers Provided, No Question Stem
Quick Reference for Question Types
This page was translated automatically. Technical details may differ from the original.
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add coffe-d/shenlun.skill