The Agent Skills Directory

[COMMAND_EXECUTION]: The skill whitelists the execution of arbitrary firecrawl and npx firecrawl commands in the shell via the allowed-tools configuration.
[COMMAND_EXECUTION]: The Firecrawl tool's -o, --output <path> flag enables writing output to any file system path, which could be exploited to overwrite system configurations or sensitive files if the agent is compromised or misled.
[EXTERNAL_DOWNLOADS]: The use of npx firecrawl facilitates the dynamic download and execution of the Firecrawl package from the public npm registry at runtime.
[DATA_EXFILTRATION]: The skill provides functionality to transmit data to arbitrary external endpoints via the --webhook-url and --email parameters, creating a channel for data exposure.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from monitored URLs and processes it through an AI judge using the --goal instruction.
Ingestion points: Content retrieved from URLs provided in --page, --scrape-urls, or --crawl-url in SKILL.md.
Boundary markers: No delimiters or protective instructions are used to isolate the ingested content from the agent's instructions.
Capability inventory: Shell command execution (firecrawl) with network access and file system write capabilities defined in SKILL.md.
Sanitization: No sanitization or validation of the monitored page content is performed before processing.

firecrawl-monitor