Workflow & Core Tools · lijigang/ljg-skills

ljg-explain

Name: ljg-explain
Author: lijigang/ljg-skills

"Explains concepts, formulas, mechanisms, and principles until they can be recognized, derived, and transferred: connecting specific scenarios, criteria for judgment, operational processes, and underlying conditions using plain Chinese. USE WHEN the user invokes ljg-explain, asks to 'explain in simple terms,' 'thoroughly explain this concept or formula,' 'Feynman technique explanation,' 'ELI5,' or wants to understand a knowledge point from scratch and apply it to new contexts. NOT FOR pure polishing or translation, full-text reading accompaniment, entire book or paper breakdowns, or requests strictly asking for formal proofs or operational steps."

L1 Basic · Source registration onlyV1 Commit-PinnedDual-pass Audit

Author lijigangGitHub lijigang/ljg-skills ↗License MIT（仓库根 LICENSE）Commit 9e75497a95

Control levels describe skill capabilities and suggested control measures; they do not represent quality or safety certifications, nor do they dictate usability. Usage decisions rest entirely with users and administrators. Different agent hosts apply varying constraints; do not assume hosts provide automatic protection. Third-party skill copyrights belong to their original authors. Use at your own risk.

1How It Works & Mechanism

The completion standard requires clarity on five things: knowing why it is needed, being able to recognize it, being able to explain the outcome, being able to adjust judgments as conditions change, and being able to transfer it; the material is interwoven into a single narrative thread.

skills/ljg-explain/SKILL.md

1. **知道为什么需要它。** 从一个真实疑惑或具体问题接到核心定义，说明它属于什么、解决什么问题。

What this is doing: replacing fixed template sections with five comprehension criteria to evaluate whether the explanation has truly hit the mark.

Audience and delivery: default to explaining for an adult with everyday life experience but lacking specialized expertise; explain directly whenever the meaning is reasonably clear without pausing to fill out a background questionnaire.

skills/ljg-explain/SKILL.md

能合理确定含义就直接讲；只有不同解释会改变讲解对象时才澄清，不能为了填写背景表而中断。

What this is doing: minimizing clarification costs and preventing background surveys from disrupting the flow of the explanation.

Default to a single complete output in the conversation, concluding naturally where the explanation wraps up, without appending self-test questions or reference answers; only split into multiple rounds when the user explicitly requests interaction.

skills/ljg-explain/SKILL.md

默认一次完整输出到对话，在讲解自然收束处结束，不在结尾追加自测题或参考答案

What this is doing: defining the delivery format: a standalone piece in a single pass that ends cleanly at the conclusion.

Formulas and principles: map symbols to concrete objects, quantities, and units; clarify what inputs enter the formula, what is computed, and explain why operations are combined in this way.

skills/ljg-explain/SKILL.md

把符号对应到具体对象、数量与单位，讲清公式输入什么、算出什么。

What this is doing: grounding formulas from abstract symbol strings back into identifiable real-world objects and operational meanings.

Chinese expression: anchor the opening sentence on the core judgment or the user's primary doubt, remove opening pleasantries and writing-process hints, and use familiar short words and tangible objects.

skills/ljg-explain/SKILL.md

第一句落在核心判断或读者真正关心的疑惑上。删去开场白、泛泛背景与写作过程提示。

What this is doing: applying opening and lexical discipline to eliminate translationese and unnecessary padding.

Evidence and formatting: explicitly state assumptions for hypothetical pedagogical examples, cite sources for real data, clearly distinguish facts, inferences, simplified models, and unknowns, and faithfully retain gaps.

skills/ljg-explain/SKILL.md

虚构的教学算例显式给出假设；引用真实数据时给来源。

What this is doing: embedding evidential rigor into the explanation to prevent examples from claiming empirical weight they do not possess.

Gotchas: fluency and paraphrasing do not prove understanding; passing content self-checks cannot be used to claim the reader has mastered the topic or to self-rate the explanation '90 out of 100.'

skills/ljg-explain/SKILL.md

检查讲解是否交代新情境下的判断依据、结果推导与条件变化；内容自审通过，也不能声称读者已经学会或给讲解自评「90 分」。

What this is doing: drawing a clear line between self-review and learning validation to prevent false-completion claims.

This skill is self-contained and completes the explanation directly: it does not automatically call another explanation skill, nor does it create notes, flashcards, reminders, or other artifacts for the explanation task.

skills/ljg-explain/SKILL.md

本技能自包含，直接完成讲解；不自动调用另一讲解技能，也不因讲解任务创建笔记、卡片、提醒或其他产物。

What this is doing: constraining skill boundaries and artifacts: delivering only an explanation text body without disk writes or delegation.

2Core Capabilities

01Explain concepts, formulas, mechanisms, and principles until they can be recognized, derived, and transferred

02Feynman-style / ELI5 plain Chinese explanations (corresponding to USE WHEN triggers like 'explain in simple terms,' 'Feynman explanation,' 'ELI5')

03Judgment criteria and contrast of easily confused ideas: prioritize minimal-difference examples that yield contrasting judgments

04Condition changes and transfer testing: modify one key condition to demonstrate how outcomes shift and when the original judgment fails

05Formulas brought to life: mapping symbols to objects and units, running full numerical examples once, and verifying formulas against units, proportions, signs, and rounding

06Discipline in native Chinese expression: core judgment first, short words for tangible objects, and reduction of mechanical conjunctions and translationese

07Self-contained without generating files; follows specified formats and locations only when explicitly requested by the user

4Recommended Governance & Evidence

Suggested Controls for Safe Operation

For Agent Runtime

Can be auto-loaded

For Administrators

Evidence & Triggers (File Paths & Excerpts)

SKILL.md: Pure prompt instructions and rule texts; no script execution or external impact

Audit Notes: Evaluation target = this skill directory (skills/ljg-explain/: SKILL.md + agents/openai.yaml). Item-by-item verification: zero scripts, zero network endpoints, zero credential reading, zero disk-write instructions ('default to not writing notes or other files'); §3 risk scan matches 0 across all seven categories. Entire behavior = generating a conversational explanation following a set of comprehension criteria and Chinese expression discipline, thus graded green.

5Second-Pass Verification

[ok] 文件构成 — §1 清单仅 SKILL.md 与 agents/openai.yaml，无脚本、无资产、无端点。
[ok] 理解标准五条自洽 — 为什么需要/能辨认/能解释结果/能随条件改判断/能迁移五条在正文与 Gotchas 中一致体现。
[ok] 交付与文件纪律 — 「默认一次完整输出到对话」「默认不写笔记或其他文件」与自包含声明一致。
[ok] agents/openai.yaml 用途 — 仅含 interface 的 display_name/short_description/default_prompt，无额外能力或端点。
[ok] 风险扫描复核 — §3 七类（URL/凭证/执行/写盘/网络/危险旗标/沙箱）均 0 命中。
[ok] 判级口径 — 纯指令、无本地写入 → 按范式落 green。

6Verdict & Scenarios

Uses five comprehension criteria (recognition, derivation, condition changes, and transfer) to turn 'thorough explanation' into verifiable standards rather than a vague claim of 'clarity.'

Maintains clear evidential boundaries: separates assumptions, real data, simplified models, and unknowns, explicitly avoiding conflating correlation with causation or fabricating stories for formulas.

Implements concrete Chinese writing discipline (core judgment first, short words for tangible objects, removal of translationese), and remains self-contained with no generated files for lightweight delivery.

Best for: Users who want to understand a concept, formula, mechanism, or principle from scratch and be able to recognize, derive, and transfer judgments in new scenarios; also well-suited for requests like 'explain in simple terms,' 'Feynman explanation,' and 'ELI5.'

Not for: Pure text polishing or translation, full-text reading accompaniment, entire book or paper breakdowns, and requests strictly requiring formal proofs or operational steps; scenarios requiring saved notes/flashcards or multi-turn interactive tutoring require explicit user requests before being handled separately.

Install Agent-Ready

① Verified Mirror 2026-10-10

Option A · Direct DownloadDownload ljg-explain.tar.gz

sha256: 8dbb7f21ea6322ac…

Option B · Copy Guide for Agent

install.json

Agents automatically download and verify mirrors from this JSON guide.

② Upstream GitHub

GitHub Repo ↗

Pinned to commit 9e75497a95.