The Agent Skills Directory

COMMAND_EXECUTION (MEDIUM): The skill relies on executing a local Node.js script (firecrawl-api.cjs) via the Bash tool. Because this file's content is not provided, the actual operations performed on the local system cannot be verified, posing a risk of undocumented behavior or vulnerabilities.
INDIRECT_PROMPT_INJECTION (LOW): This skill is vulnerable to instructions hidden within scraped web pages.
Ingestion points: Data enters the system via the url parameter used in scrape, crawl, and map functions.
Boundary markers: Absent. There are no instructions or delimiters defined to help the agent distinguish between the scraped data and its own system instructions.
Capability inventory: The skill has access to powerful tools including Bash, Write, and Task, which could be exploited if the agent follows malicious instructions found on a scraped website.
Sanitization: Absent. The content retrieved from the web is passed directly into the agent's context as markdown or HTML without filtering or escaping.

firecrawl-scraper