CtrlK
BlogDocsLog inGet started
Tessl Logo

yj-ocr-parser

文档解析技能,用于解析PDF文档和图片(jpg/png/jpeg)并将其转换为Markdown格式输出。支持标题层级、表格(HTML格式)、公式(LaTeX格式)、图片(链接形式)等复杂内容的解析。当用户提到"解析PDF"、"文档解析"、"PDF转Markdown"、"提取PDF内容"、"解析文档"、"文档内容提取"、"PDF内容识别"、"图片解析"、"图片转文字"、"识别图片内容"等场景时使用此技能。即使用户只是要求读取或查看PDF/图片文件内容,也应考虑使用此技能。

70

Quality

85%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

High

Do not use without reviewing

SKILL.md
Quality
Evals
Security

Quality

Content

75%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.

Highly actionable content: a real endpoint, executable curl commands, a parameter table, the response schema, and explicit error-code handling make this copy-paste ready. The main weaknesses are redundancy — token instructions repeated three times, the workflow listed twice, and near-duplicate examples — which inflates length without adding information.

Suggestions

Collapse "使用步骤" and "执行流程" into a single authoritative step list, keeping the error-code validation from step 6 of 执行流程.

State the MaaS_model_token sourcing rule once (in 变量要求) and remove the repeated explanations in the curl note and the "认证方式" section.

Drop one of the two near-identical example curl commands (the PDF and image examples differ only in path/filename) and keep a single example with a note that image files are passed the same way.

DimensionReasoningScore

Conciseness

The content is mostly skill-specific and efficient, but includes clear padding that could be tightened: the MaaS_model_token sourcing instruction is repeated three times ("变量要求", the note after the curl block, and "认证方式"), the workflow steps are listed twice ("使用步骤" and "执行流程"), and the two example curl commands are near-duplicates of the template. It is above anchor 2 because none of it explains concepts Claude already knows — the redundancy is restatement, not generic filler.

3 / 5

Actionability

The curl command is fully executable with a real endpoint, a parameter table ("file | 是 | multipart file", "file_name | 是 | string"), the exact success response JSON schema, per-code error handling ("code 为 '429' 表示令牌限流"), and two copy-paste-ready examples with concrete paths. This matches the top anchor: complete commands with specific examples covering the common cases.

5 / 5

Workflow Clarity

"执行流程" provides a clear 7-step sequence with real checkpoints: verify file existence and format (steps 1-2), verify the token is present (step 3), and inspect the response code with per-code actions (step 6). It falls short of anchor 5 because the error recovery is user-facing messaging ("提示用户稍后重试") rather than an explicit validate-fix-retry loop, and the parallel "使用步骤" list partially duplicates and muddies the canonical sequence.

4 / 5

Progressive Disclosure

The skill has no bundle files and is a single-API-call task, and the body is organized into well-labeled sections (适用场景, 变量要求, API 调用方式, 参数说明, 执行流程, 返回结果处理, 注意事项, 示例) that are easy to navigate. It is not 5 because at ~113 lines it exceeds the lean single-screen overview the simple-skill exception rewards — the parameter table and duplicated examples could be trimmed or split — but there is no misplacement or buried-reference problem, so 3 does not fit.

4 / 5

Total

16

/

20

Passed

Description

96%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 states concrete capabilities with format-level detail, provides an explicit and comprehensive set of natural trigger phrases, and clearly answers both what the skill does and when to use it. The only weakness is the deliberate broadening to generic 'read/view PDF' requests, which slightly increases conflict risk with general file-handling skills.

DimensionReasoningScore

Specificity

The description lists multiple concrete actions — "解析PDF文档和图片(jpg/png/jpeg)并将其转换为Markdown格式输出" with explicit handling of "标题层级、表格(HTML格式)、公式(LaTeX格式)、图片(链接形式)" — covering the capability surface comprehensively. It exceeds anchor 4 because there are no meaningful gaps in the actions described, not just 'several with minor gaps'.

5 / 5

Completeness

Both 'what' (parse PDFs/images into Markdown with headings, HTML tables, LaTeX formulas, image links) and 'when' ("当用户提到…等场景时使用此技能" with concrete trigger phrases, plus "即使用户只是要求读取或查看…也应考虑使用") are explicitly and clearly answered. This mirrors the anchor-5 example structure exactly, so 4 ('when could be more explicit') does not fit.

5 / 5

Trigger Term Quality

It enumerates natural user phrases users would actually say — "解析PDF", "文档解析", "PDF转Markdown", "提取PDF内容", "图片解析", "图片转文字", "识别图片内容" — including synonyms and file extensions (jpg/png/jpeg). This matches the comprehensive synonym-plus-extension coverage of the top anchor; anchor 4 ('a few natural terms missing') understates the breadth present.

5 / 5

Distinctiveness Conflict Risk

The core niche (PDF/image → Markdown parsing via a specific API) is distinct with clear triggers, but the closing clause "即使用户只是要求读取或查看PDF/图片文件内容,也应考虑使用此技能" broadens triggering to generic read/view requests, creating minor overlap risk with general file-reading skills. It is not 5 because of that broadened trigger; it is clearly above 3 since the domain and triggers are specific rather than merely 'somewhat specific'.

4 / 5

Total

19

/

20

Passed

Validation

100%

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

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
UnicomAI/wanwu
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.