The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill requires installation of firecrawl-cli@1.4.1 from npm (rules/install.md), which is an external dependency from a source not on the pre-approved trusted list.
[PROMPT_INJECTION] (LOW): High surface area for Indirect Prompt Injection (Category 8) due to its core function of scraping untrusted web data.
Ingestion points: Untrusted data enters the agent context via scrape, search, crawl, and agent commands defined in SKILL.md.
Boundary markers: The skill documentation (SKILL.md) explicitly instructs the agent to isolate content using the -o flag to write output to a .firecrawl/ directory, preventing direct injection into the instruction prompt.
Capability inventory: The skill utilizes shell execution for CLI operations and network access for web data retrieval.
Sanitization: The documentation recommends using incremental reading tools like grep and head to minimize exposure to potentially malicious instructions in scraped content.
[COMMAND_EXECUTION] (SAFE): Commands are well-defined and include safety advice to quote URLs to prevent command injection.
[CREDENTIALS_UNSAFE] (SAFE): Authentication instructions in rules/install.md use placeholders for API keys and standard OAuth flows.

firecrawl