# Search Aozora Bunko and Extract Literary Text

Finding Japanese literary texts requires navigating catalog records and inconsistent page encodings. This skill searches the official catalog and extracts readable text from linked works.

## Install

```bash
npx skillstore add tahodev/aozora
```

## Metadata

- Status: approved
- Slug: tahodev-aozora
- Skillstore revision: r1
- Version status: missing
- Tree hash: 88c47511945aa580ae66b76263568afc16c5c3e329bcb69e1232ec71366db28f
- Author: tahodev
- GitHub username: tahodev
- License: MIT
- Repository: https://github.com/tahodev/kurashi-skill/tree/09cb7f9eac85f0c85633cce08328a1a57854a94f/aozora
- Ref: 7f17813fb6af1acc7d8fd3fb5f05480f5be52783
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, filesystem, external\_commands
- Quality score: 73
- Public page: https://skillstore.pages.dev/skills/tahodev-aozora
- Manifest: https://skillstore.pages.dev/api/skills/tahodev-aozora/manifest

## Capabilities

- Search titles by substring and authors by exact concatenated name, with configurable result limits and work-level deduplication.
- Return work IDs, titles, kana readings, authors, orthography, copyright flags, first-publication details, and card URLs for search results.
- Resolve work IDs through catalog records and extract XHTML, ruby ZIP, and ebook links from book-card pages.
- Decode Shift\_JIS or UTF-8 XHTML and extract plain text while removing ruby readings and annotation parentheses.
- Report catalog rows, distinct works, distinct people, and copyright-flagged work counts.
- Cache the official catalog for 24 hours or search a supplied local CSV; limit returned text by character count.

## Use Cases

- Locate Literary Reading Material: Find works by title or author and inspect readings, orthography, copyright flags, and links before choosing classroom material.
- Inspect Catalog Records: Retrieve work identifiers, publication details, and contributor roles from official catalog records and linked cards.
- Prepare Text Samples: Extract bounded plain-text samples without ruby readings from selected works after checking rights and source provenance.

## Prompt Templates

### Find a Title

```
Search Aozora Bunko for 吾輩は猫である. Show the work ID, author, kana reading, copyright flag, and card URL.
```

### List an Author's Works

```
Find up to ten works by 夏目漱石. Show titles, work IDs, orthography, and copyright flags, then report the total matching works.
```

### Retrieve a Text Excerpt

```
Check copyright status for work 000789, resolve its card links, and retrieve the first 500 plain-text characters without ruby readings.
```

### Inspect a Local Catalog

```
Use my trusted local Aozora CSV without network access. Report row, work, person, and copyright-flagged counts; explain duplicate contributor rows.
```

## Limitations

- The catalog is a periodically updated snapshot, not a live inventory; some cards have no available text.
- Title spelling must match stored characters, and author searches require exact concatenated names without spaces.
- Copyrighted works are included, and retrieval does not enforce public-domain status or establish reuse permission.
- Fetched URLs and download sizes lack security limits; character limits only shorten output after the full text is downloaded.

## Best Practices

- Check each work's copyright flag and applicable permissions before reusing or distributing text.
- Use trusted catalog data and restrict network destinations when running the helper.
- Use exact stored names, preserve six-digit work IDs, and choose character limits for manageable excerpts.

## Anti Patterns

- Treating every catalog entry as public domain or assuming online access grants unrestricted reuse.
- Guessing text URLs from work IDs instead of reading the book-card links.
- Using catalog row counts as work counts or treating snapshot statistics as current live totals.

## Security Audit

- Audited at: 2026-10-13T03:07:27.522\+00:00
- Summary: Most alerts are harmless Markdown, Japanese text, fixture references, or scoped cache maintenance. Confirmed risks involve unchecked fetch destinations, redirects, and unbounded download or ZIP expansion sizes. No evidence found of prompt injection or credential exfiltration, but the advertised public-domain scope is not enforced.

## Stats

- Views: 0
- Downloads: 0
- Favorites: 0
- Popularity score: 0