The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted content from the web, academic databases (OpenAlex, PubMed), and metadata repositories (OAI-PMH). If an agent processes this scraped text without sanitization, it could be susceptible to instructions hidden in the data.
Ingestion points: Scraping functions in academic-web-scraping/SKILL.md, Google Scholar results in google-scholar-scraper/SKILL.md, and dataset metadata in dataset-finder-guide/SKILL.md.
Boundary markers: None identified in the provided extraction logic to distinguish data from instructions.
Capability inventory: The skill has file system write access (DataCollector class), network request capabilities (requests, urllib), and browser automation via Playwright.
Sanitization: No sanitization or filtering of external content is demonstrated in the code examples.
[EXTERNAL_DOWNLOADS]: The skill facilitates interaction with various external research platforms and tools.
Fetches scholarly data and metadata from well-known services including OpenAlex, Crossref, PubMed (NCBI), and Zenodo.
easy-spider-guide/SKILL.md provides instructions to download and install the EasySpider visual crawler from its GitHub repository.
dataset-finder-guide/SKILL.md includes a reproducible download script that verifies data integrity using SHA-256 checksums, which is a security best practice.
[COMMAND_EXECUTION]: The documentation includes shell commands for environment setup and tool usage.
easy-spider-guide/SKILL.md contains commands for cloning repositories, installing Node.js dependencies, and executing AppImages (chmod +x).
dataset-finder-guide/SKILL.md provides instructions for using the Kaggle CLI tool for dataset management.

scraping-skills