jp-report

Create formal Japanese corporate HTML and PDF reports from source text for business, compliance, or security documents with structured chapters and professio...

Security Scan

The skill is internally consistent with its stated purpose (generate Japanese corporate HTML + PDF), requests no credentials, has no network endpoints, and its only code runs a local headless Chrome to render PDFs.

Purpose & Capability

Name/description (formal Japanese reports -> HTML+PDF) match the included files: docs for CSS/components, example HTML, and a small script to call headless Chrome. The macOS/Chrome requirement declared in README explains the script's hardcoded Chrome path.

Instruction Scope

SKILL.md restricts operations to reading the provided source text (file or pasted content), building HTML using local docs, and invoking the included Python script to produce a PDF. It references ${CLAUDE_SKILL_DIR} (platform-provided skill directory) to embed local files; this variable is not declared in requires.env but is a normal platform runtime variable. No instructions read unrelated system files or contact external endpoints.

Install Mechanism

No install spec (instruction-only skill) and only a tiny Python script is included. Nothing is downloaded or executed from remote URLs. The README's manual symlink installation is benign guidance.

Credentials

The skill declares no required environment variables, no credentials, and reads only user-provided input files and local docs/components for HTML/CSS. There are no secrets or unrelated env vars requested.

Persistence & Privilege

always:false and the skill does not request persistent system privileges or modify other skills. SKILL.md sets disable-model-invocation:true which reduces the skill's ability to autonomously call the model (not a risk).

Assessment

This skill appears coherent and limited to generating paginated HTML and producing a PDF via a local headless Chrome call. Before installing, consider: 1) The generate_pdf.py script invokes a hardcoded macOS Chrome binary path (/Applications/Google Chrome.app/Contents/MacOS/Google Chrome) — ensure you are on macOS with Chrome installed at that location or update the script. 2) The script URL-encodes the path using urllib.parse.quote on the entire path (could produce incorrect file:// URLs for some systems); test PDF generation with filenames containing spaces and Japanese characters. 3) The skill uses ${CLAUDE_SKILL_DIR} to read local docs (design-rules.md, components.md) — this is expected but platform-specific. 4) The skill writes .html/.pdf to whatever output folder the user supplies — do not point it at sensitive system directories. 5) No network calls or credentials are requested, so there's no obvious data exfiltration vector in the files provided. If you need to be extra cautious, review or run the generate_pdf.py locally to confirm it behaves as expected before granting it access to production documents.

Generate a formal Japanese corporate document (日本企業向け正式報告書) from the provided source material.

INPUTS

If $ARGUMENTS is provided, treat it as the source file path or inline content. Otherwise, ask the user for:

1. Source material — file path or pasted content
2. Document purpose — e.g. security overview / compliance / product explanation
3. Japanese title + English subtitle
4. Classification level — 社外秘 / 部外秘 / 社内限り (default: 社外秘)
5. Output folder path
6. Revision number (default: Rev. 1.0)

---

STEP 1 — ANALYZE

Read source material. Identify:

• Logical sections → map to 4–6 chapters
• Content types: principles, tables, architecture diagrams, bullet lists
• Brand names / technical terms to preserve verbatim
• Content to exclude (confirm with user if unclear)

---

STEP 2 — TRANSLATE TO FORMAL JAPANESE

Rewrite all content in 書き言葉・敬体:

• Verb endings: 〜しております / 〜いたします / 〜となっております
• Full grammatical sentences — no fragment bullets inside tables or definitions
• Preserve intact: brand names, acronyms (RBAC, TLS, SSO, BYOK), product names
• Do NOT add certifications (SOC/ISO/GDPR) unless explicitly in source

---

STEP 3 — PLAN PAGES

Fixed structure:

Page 1  —  表紙 (Cover)   : centered layout, one full A4 page
Page 2  —  目次 (TOC)     : dotted leaders + page numbers, one full A4 page
Page 3+ —  本文 (Body)    : chapters distributed, target one chapter per page

Numbering: 第１章 / 第２章 … (full-width) · Sections: 1.1 / 1.2 … (decimal)

Before writing HTML, calculate content height using the reference in ${CLAUDE_SKILL_DIR}/docs/design-rules.md. Usable area per page ≈ 980px. State the page plan briefly to the user.

---

STEP 4 — GENERATE HTML

Write a single self-contained .html file using:

• CSS from ${CLAUDE_SKILL_DIR}/docs/design-rules.md (include verbatim in <style>)
• Component HTML from ${CLAUDE_SKILL_DIR}/docs/components.md

Every page must follow this shell:

<div class="page">
  <div class="page-body"> <!-- content --> </div>
  <div class="page-footer">
    <div class="f-l">[Product / Company]</div>
    <div class="f-c">[Document title]　Rev. X.X</div>
    <div class="f-r">[Classification]　｜　[Page #]</div>
  </div>
</div>

Cover page body must use:

<div class="page-body" style="display:flex;flex-direction:column;align-items:center;justify-content:center;text-align:center;">

---

STEP 5 — GENERATE PDF

Immediately after writing the HTML, run:

python3 ${CLAUDE_SKILL_DIR}/scripts/generate_pdf.py [absolute path to .html]

The script handles Chrome invocation and URL-encodes the path automatically. Report the output line (bytes + path) on success.

---

QUALITY CHECKLIST

• Cover: content vertically + horizontally centered
• TOC: page numbers match actual layout
• Every .page div has a .page-footer
• No bright colors, emoji, or casual language
• PDF generated successfully (bytes confirmed)