The Agent Skills Directory

[SAFE]: The skill provides instructions for a legitimate web crawling utility. External resources (fastcrw.com, github.com/us/crw, $CRW_API_URL) are verified as vendor-owned by 'us' and do not involve typosquatting or suspicious redirection.\n- [SAFE]: The skill implements the 'allowed-tools' frontmatter field, which is a platform-level configuration that limits the agent's available tools to specific Bash and Read operations, following the principle of least privilege.\n- [SAFE]: Best practices for safety and cost-efficiency are encouraged, such as using 'crw-map' first to estimate scope, starting with conservative crawl limits (depth 1, limit 10), and using aggressive rate-limiting for production sites.\n- [SAFE]: Data management instructions recommend saving crawl outputs to a local directory ('.crw/') and processing files incrementally rather than overloading the model's context window with large data blobs.

crw-crawl