skills/xberg-io/plugins/crawlberg/Gen Agent Trust Hub

crawlberg

Pass

Audited by Gen Agent Trust Hub on Aug 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill operates by shelling out to a local crawlberg binary to perform web scraping and crawling tasks. Subcommands include scrape, crawl, map, interact, and serve.
  • [EXTERNAL_DOWNLOADS]: Installation instructions reference external vendor-managed sources across multiple package managers.
  • Homebrew tap: xberg-io/tap/crawlberg.
  • Node package: @xberg-io/crawlberg-cli.
  • Python/UV package: crawlberg-cli.
  • Rust/Cargo: crawlberg-cli.
  • [DATA_EXFILTRATION]: The tool is designed to send HTTP requests to external URLs to fetch content, which involves communication with arbitrary domains. It supports custom proxies and user-agents for these requests.
  • [REMOTE_CODE_EXECUTION]: The interact subcommand provides the ability to execute JavaScript (executeJs) within a headless browser instance, allowing for dynamic interaction with target websites.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from the web, creating a surface for indirect prompt injection where instructions embedded in scraped content could influence agent behavior.
  • Ingestion points: External web content fetched via scrape, crawl, map, and batch-scrape subcommands.
  • Boundary markers: The documentation does not explicitly define delimiters for scraped content; it relies on conversion to Markdown.
  • Capability inventory: The tool set includes subprocess execution (CLI), network operations, and browser automation.
  • Sanitization: Content is processed through a built-in HTML-to-Markdown engine to generate clean output.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 22, 2026, 10:23 PM
Security Audit — agent-trust-hub — crawlberg