pdf-explore

Use this skill when the user has attached a PDF, paper, report, or other document and the answer needs content from more than one place in it: summarize the methods or any other section, compare sections, find where a topic is discussed, read a value or label off a figure or chart, or find/list/extract every instance of something across the whole document (datasets, benchmarks, citations, figures, table rows, accession numbers — including appendices). Parses the PDF once with a deterministic Python kernel: `pdf_pages` (pages as persistent text, or high-res images), `pdf_outline` (embedded TOC), `pdf_scan` (a lexical pre-filter that narrows a long doc to candidate pages), `pdf_grep` (regex sweep for exhaustive pattern extraction). You read the shortlist and do the relevance / summary / extraction judgment yourself. For PDF creation/manipulation, use reportlab/pypdf directly.

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

77%

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

A highly actionable, clearly sequenced skill body with executable recipes and good workflow guidance. The main weaknesses are some repetition and a license digression that hurt conciseness, and a monolithic structure with no progressive disclosure despite the skill's size.

Suggestions

Move the helper API reference table and/or the license/backend notes into a short REFERENCE.md and link to it one level deep, slimming the main body.

Trim the repeated token-cost comparisons and the AGPL licensing paragraph unless a user is likely to embed PyMuPDF; keep one concise cost statement.

The kernel.py referenced in Setup is not present in any bundle directory — either ship it or note its location so the exec line is verifiable.

Dimension	Reasoning	Score
Conciseness	Mostly efficient with executable recipes and a clear helper table, but it restates the token-cost point across several sections and includes a license digression (PyMuPDF AGPL terms) and stride-sampling caveats that are not strictly needed to use the skill, fitting the anchor-2 'could be tightened'.	2 / 3
Actionability	Every recipe is complete, executable Python with concrete helper signatures, return shapes, and copy-paste-ready snippets (e.g., the pdf_pages write-to-file loop and the pillow crop), matching the fully-executable anchor.	3 / 3
Workflow Clarity	Recipes are explicitly sequenced with a 'try this first' ordering and clear feedback guidance ('broaden top_k if the shortlist looks thin', 'decide all your doubts up front and fetch them in one call'), giving clear sequences with error-recovery checkpoints as the anchor-3 example does.	3 / 3
Progressive Disclosure	No bundle files exist (references/, scripts/, assets/ all absent) and the body is a single ~240-line monolith with the helper reference table and all recipes inline; it is well-sectioned but content that could be split into one-level-deep references is not, fitting the anchor-2 'content that should be separate is inline'.	2 / 3
	Total	10 / 12 Passed

Description

90%

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 highly specific, well-triggered description with explicit when/what guidance and a clear niche. The only deduction is for second-person voice, which the rubric penalizes on specificity.

Suggestions

Rewrite "You read the shortlist and do the relevance / summary / extraction judgment yourself" in third person (e.g., "The agent reads the shortlist and applies the relevance/summary/extraction judgment") to avoid the voice penalty.

The helper enumeration (pdf_pages, pdf_outline, pdf_scan, pdf_grep) is implementation detail that pads the description; consider trimming it so the trigger and intent stay foregrounded.

Dimension	Reasoning	Score
Specificity	Lists many concrete actions and named helpers (pdf_pages, pdf_outline, pdf_scan, pdf_grep), matching the anchor-3 specificity, but the second-person phrasing "You read the shortlist and do the relevance / summary / extraction judgment yourself" triggers the rubric's voice penalty, reducing the score by one.	2 / 3
Completeness	Explicitly opens with "Use this skill when the user has attached a PDF..." and then enumerates what it does, clearly answering both what and when as the anchor-3 example does.	3 / 3
Trigger Term Quality	Strong coverage of natural user terms ("PDF", "paper", "report", "summarize", "compare sections", "figure", "chart", "datasets", "benchmarks", "citations", "appendices") that users would actually say, matching the good-coverage anchor.	3 / 3
Distinctiveness Conflict Risk	Has a clear niche (reading content from multiple places in an already-attached PDF) and an explicit disambiguator — "For PDF creation/manipulation, use reportlab/pypdf directly" — making conflict with adjacent skills unlikely.	3 / 3
	Total	11 / 12 Passed

Validation

100%

Warnings & errors only

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
Path: configs/microservice/bff-service/configs/agent-skills/claude-science/pdf-explore/SKILL.md
Commit: f618458

Reviewed: about 14 hours ago

Table of Contents

Discovery Implementation Validation

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.