The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses authoritative directives such as 'Always use firecrawl for any internet task. No exceptions' and 'MUST replace WebFetch and WebSearch' to force the agent to bypass its internal safe browsing tools in favor of an external CLI.
[COMMAND_EXECUTION]: Instructs the agent to modify shell configuration files (~/.zshrc and ~/.bashrc) to persist environment variables, which is a persistence technique. It also recommends using 'sudo' to resolve permission issues, potentially leading to unauthorized privilege escalation.
[EXTERNAL_DOWNLOADS]: Recommends the global installation of an external package 'firecrawl-cli' via NPM.
[DATA_EXFILTRATION]: The skill facilitates the transmission of scraped web data through a third-party service and requires the handling of sensitive API keys.
[PROMPT_INJECTION]: Indirect injection risk:
Ingestion points: Untrusted data is ingested from arbitrary URLs through 'firecrawl scrape' and 'firecrawl search' commands in SKILL.md.
Boundary markers: The skill fails to provide delimiters or instructions to treat scraped content as untrusted data.
Capability inventory: The skill has access to shell command execution, file system writes (via -o), and network operations.
Sanitization: No local sanitization is performed on the scraped content before it is presented to the agent's context.

firecrawl