CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

ljg-book

Explain a whole book to readers without specialist knowledge: what it follows, how its main threads connect, where it ends, and how its evidence or events change understanding. Keep the book's specific content; a general lesson alone is not enough. USE WHEN given a book title, PDF, EPUB, excerpt, or asked 拆书, 分析这本书, 这本书在讲什么, 压缩一本书, or book. Defaults to a saved Org note. NOT FOR chapter summaries, papers, formal reviews, or single-concept deep dives.

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

Quality

Content

63%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body delivers a well-differentiated writing workflow with unusually concrete output contracts (filename pattern, save location, validator, Emacs verification) and a real evaluation feedback loop. Its weaknesses are the dangling must-read reference to a missing ReadingGuide.md, duplicated 末节 guidance, and abstract prose in the middle sections that could be condensed without losing instruction.

Suggestions

Add the missing ReadingGuide.md to the bundle (it is marked 必读 in the routing table and cited in Completion) or path-qualify and fix the reference, e.g. `references/ReadingGuide.md`, so the primary required reading actually resolves like `references/template.org` does.

Deduplicate the ending-section guidance: the Gotchas bullet "末节只收拢已经形成的理解,不重复开头..." and the section "末节:读完后应发生的变化" repeat the same three points nearly verbatim — keep the detailed section and reduce the Gotcha to a one-line pointer.

Tighten the abstract middle sections ("让一个念头有机会走完", "用问题的缺口拉动阅读") by consolidating restatements of the same principle (e.g., the multiple formulations of "don't extract the lesson too early") into a single concrete instruction with one worked example.

DimensionReasoningScore

Conciseness

The body mostly avoids explaining things Claude already knows and packs real instruction into each line, but it could be tightened: the 末节 guidance appears twice in near-verbatim form (the Gotcha "末节只收拢已经形成的理解,不重复开头...不需要重新建立问题" versus the "末节:读完后应发生的变化" section repeating "末节的任务是收获,不是重复开头...不需要重新建立问题的张力"), and the middle sections ("让一个念头有机会走完", "用问题的缺口拉动阅读") spend tokens on evocative abstract prose ("好奇心需要一个空隙才能产生") that restates the same principle in multiple ways. This matches anchor 3 — mostly efficient with some tightening possible — rather than anchor 4's minor trims.

3 / 5

Actionability

For an instruction-only writing skill, the guidance is mostly concrete: a routing table mapping each input type to required reading and output, an exact filename pattern "{YYYYMMDDTHHMMSS}--拆书-{书名}__book.org" with the generating command "date +%Y%m%dT%H%M%S", a hard constraint ("每个图块或表格不超过 80 显示列"), a four-item numbered breakpoint checklist, the validator invocation "bun {skill_dir}/scripts/validate_note.ts ... --coverage ...", and named Emacs verification functions. It stops short of anchor 5 because much of the core writing instruction (e.g., "让原来的判断先有成立的理由", "术语在需要它之前不出现") is directional principle rather than operationalized guidance with worked examples, leaving minor gaps.

4 / 5

Workflow Clarity

The sequence is present and validated: the routing table gives entry steps per input type, the evaluation section defines an explicit feedback loop ("发现断点就修订相应内容,再读受影响的连接" — find breakpoints, revise, re-read affected connections), and Completion acts as a checklist with the validator, per-warning read-back ("逐条读回 warning"), and honest-reporting requirements ("不可用时明确延期,不能把未执行写成通过"). It does not reach anchor 5 because the full pipeline (read guide → analyze → write → evaluate → revise → validate → Emacs verify) is never laid out as one coherent ordered flow — it is distributed across thematically organized sections the reader must assemble, and "ReadingGuide.md" is the designated first read yet its content/availability is not established.

4 / 5

Progressive Disclosure

Structure exists — a routing table signaling which file to read per input ("先读原文,再读 `ReadingGuide.md`", "写 Org 时读 `references/template.org`", "复制 `references/coverage-map.md`"), one-level-deep references, and genuinely substantive bundle files (template.org, coverage-map.md, validate_note.ts all verified present). However "ReadingGuide.md" is cited twice as 必读/must-read (routing table and Completion: "细则见 `ReadingGuide.md`") but does not exist anywhere in the bundle, and it is referenced without a path prefix unlike the references/ files — a broken, unclearly-signaled reference to the skill's primary reading guide. This matches anchor 3 (references present but not clearly signaled) rather than anchor 4's minor organization gaps.

3 / 5

Total

14

/

20

Passed

Description

95%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description: it explicitly states what the skill produces, when to trigger it with natural multilingual phrases and file formats, and what to exclude. The only soft spot is that the capability is expressed as facets of a single explanatory task rather than a list of distinct concrete actions, which keeps specificity at 4 rather than 5.

DimensionReasoningScore

Specificity

The description lists several concrete facets of the deliverable — "what it follows, how its main threads connect, where it ends, and how its evidence or events change understanding" plus "Keep the book's specific content" — which goes beyond naming the domain. It falls short of anchor 5 because these are aspects of one composite action (explain a whole book) rather than multiple distinct concrete actions, and output specifics like the Org-note format are only implied by "Defaults to a saved Org note".

4 / 5

Completeness

Both halves are explicit: what it does ("Explain a whole book to readers without specialist knowledge... how its evidence or events change understanding") and when to use it ("USE WHEN given a book title, PDF, EPUB, excerpt, or asked 拆书...") with concrete trigger phrases. It even adds boundary exclusion ("NOT FOR chapter summaries, papers, formal reviews, or single-concept deep dives"), exceeding the anchor-5 example, and it uses third person throughout, so no voice penalty applies.

5 / 5

Trigger Term Quality

"USE WHEN given a book title, PDF, EPUB, excerpt, or asked 拆书, 分析这本书, 这本书在讲什么, 压缩一本书, or book" comprehensively covers natural phrasings users would actually say, including Chinese colloquial triggers (拆书, 压缩一本书) and input formats (PDF, EPUB, excerpt). This matches the anchor-5 example's coverage of synonyms and file formats; no common variation is obviously missing.

5 / 5

Distinctiveness Conflict Risk

The niche is clear (whole-book explanation preserving specific content) with distinct triggers, and the explicit "NOT FOR chapter summaries, papers, formal reviews, or single-concept deep dives" carve-out actively prevents overlap with adjacent summarization/review skills. Conflict risk is minimal, matching anchor 5.

5 / 5

Total

19

/

20

Passed

Validation

93%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

Total

15

/

16

Passed

Repository

lijigang/ljg-skills

Path

skills/ljg-book/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.