agent-readable
github.com/yamadashy/agent-readable- Category
- AI Agents
- Rank
- No. 1808Tools index
Previous survey · No. 1813 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- yamadashy
- GitHub
- 2 stars
- Date
About
Yamadashy's Agent Skills for reading hard-to-access content — gives agents the ability to pull text out of PDFs, images, and obscure formats.
What it does
A collection of agent playbooks for retrieving specific sources through sanctioned routes. Each skill recognizes a source, chooses an access method, and guides the agent through focused retrieval, transformation, or section selection.
Why it's ranked here
Worth considering for its disciplined source handling and practical orchestration. It covers seven useful targets, documents attribution and access policies, and avoids sources that prohibit automation. Its narrower value is routing and procedure, not a standalone universal reader.
What's good
The collection prefers official APIs and raw source repositories over scraping. PDF handling detects image-heavy pages and can render selected pages for visual reading. Large specifications are fetched by section, while GitHub guidance routes different URL shapes to focused command sheets.
Tradeoffs
Several workflows depend on external tools. PDF extraction requires pdfvision and a recent Node.js runtime, while GitHub access requires the GitHub CLI and authentication. Anonymous Stack Overflow access has a daily quota. The source list is intentionally narrow, and policy restrictions exclude several popular sites.
How to use it well
Best for agents that regularly research papers, web standards, GitHub activity, MDN documentation, Stack Overflow answers, PDFs, or SpeakerDeck slides. Add it as a retrieval layer before analysis. Pair it with summarization or synthesis tooling, because contributors are explicitly told to document reading procedures only.
Technical notes+
README.md installs the collection with npx skills add yamadashy/agent-readable and lists seven skills. skills/readable-pdf/SKILL.md delegates extraction to pdfvision, supports Markdown, JSON, XML, page subsets, density reporting, and PNG rendering; it requires Node.js >= 22.13. skills/readable-github/SKILL.md uses authenticated gh REST and GraphQL access, then routes URL shapes into topic references such as skills/readable-github/references/code.md. skills/readable-mdn/SKILL.md maps documentation URLs to raw GitHub Markdown and falls back to WebFetch on a 404. CONTRIBUTING.md requires single user-supplied URLs, official APIs when available, rate-limit compliance, and source-specific attribution.
Observed
- License
- MIT License
- Packaging
- Agent Skills collection installed through npx skills add
- Interfaces
- Instruction-driven CLI, WebFetch, REST API, and GraphQL workflows
- Skill coverage
- Seven listed skills covering arXiv, SpeakerDeck, PDF, GitHub, Stack Overflow, web specifications, and MDN
- PDF runtime
- pdfvision requires Node.js >= 22.13
- GitHub dependency
- GitHub workflows require the gh CLI and authentication
- Repository structure
- Each source uses a Markdown skill definition, with optional reference documents for URL routing
Read from README.md, LICENSE, CLAUDE.md, CONTRIBUTING.md, skills/readable-mdn/SKILL.md, skills/readable-pdf/SKILL.md, skills/readable-spec/SKILL.md, skills/readable-arxiv/SKILL.md, skills/readable-github/SKILL.md, skills/readable-speakerdeck/SKILL.md, skills/readable-stackoverflow/SKILL.md, .claude/commands/git/git-commit.md, .claude/skills/agent-browser/SKILL.md, .claude/commands/git/git-commit-push.md, skills/readable-github/references/code.md.
What it can do
Extract text from PDF files
PDF file → Plain text content
Extract text from images
Image file (JPG, PNG, etc.) → Text content from image
Parse obscure file formats
Files in uncommon or proprietary formats → Readable text content
Convert hard-to-read documents to accessible text
Various document formats → Agent-readable text format
Process scanned documents
Scanned document files → Extracted text content
Tags
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.