The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the firecrawl-cli package from the official npm registry and executing it via npx. These are standard installation procedures for a well-known web scraping service.
[SAFE]: Analysis of potential indirect prompt injection vulnerabilities confirms the inclusion of significant safety mitigations:
Ingestion points: External content is ingested through the scrape, search, map, and crawl commands across SKILL.md and referenced rules.
Boundary markers: The rules/security.md file mandates the use of the -o flag to isolate all fetched content within a specific .firecrawl/ directory, preventing untrusted data from directly entering the primary context window.
Capability inventory: The skill uses Bash to invoke the Firecrawl CLI, which performs network operations and local filesystem writes limited to the output directory.
Sanitization: The documentation explicitly instructs the agent to treat all fetched content as untrusted and ignore any instructions found within the data, effectively preventing indirect prompt injection from influencing agent behavior.

firecrawl