The Agent Skills Directory

[SAFE]: The skill is strictly limited to architectural design and system specifications. It explicitly prohibits the generation or execution of code, routing implementation tasks to specialized agents (e.g., 'Builder' or 'Vector'). This architectural focus prevents the risk of unauthorized system manipulation or direct remote code execution.
[SAFE]: Comprehensive compliance frameworks are integrated as structural requirements. The skill provides detailed guidance on respecting robots.txt (RFC 9309), the EU AI Act, and various machine-readable opt-out protocols (e.g., TDMRep, ai.txt). It includes specific warnings regarding legal precedents and prohibited activities like CAPTCHA circumvention.
[SAFE]: Anti-detection and infrastructure techniques are framed within an ethical and legal context. The skill requires documented authorization for crawling and mandates that measures be proportional and transparent, explicitly avoiding DDoS-equivalent traffic or stealthy patterns without a legal basis.
[SAFE]: All referenced software components (such as Scrapy, Redis, Kafka, and Playwright) and deduplication algorithms (SimHash, MinHash) are legitimate industry standards for data engineering and distributed systems. No unverified or suspicious third-party dependencies were found.
[SAFE]: The instruction set includes robust operational boundaries and 'Never' rules that enforce safety, such as prohibiting the bypassing of robots.txt directives and requiring human review for ambiguous compliance scenarios.

trawl