The Agent Skills Directory

Unverifiable Dependencies & Remote Code Execution (SAFE): The skill installs the firecrawl-cli package via npm using a pinned version (1.4.1) in rules/install.md. This is a standard and acceptable practice for CLI-based skills.\n- Privilege Escalation (SAFE): The installation instructions in rules/install.md explicitly recommend against using sudo. It provides a safer alternative by configuring a local npm global directory (~/.npm-global), demonstrating good security hygiene.\n- Indirect Prompt Injection (SAFE): The skill correctly identifies that web content fetched via scraping is untrusted. rules/security.md provides mandatory guidance for the agent, including file-based isolation in a .firecrawl/ directory and incremental reading (using grep or head) to minimize exposure to potentially malicious instructions embedded in web pages.\n- Data Exposure & Exfiltration (SAFE): No evidence of credential exfiltration or access to sensitive local files was found. The tool naturally communicates with its own service (firecrawl.dev) for its core functionality as a web scraper.

firecrawl