The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several shell commands including whois, dig, and grep. It uses environment variables (SECSKILL_TARGET) as inputs for these commands. While variables are quoted (e.g., whois "$TARGET"), providing arbitrary strings to shell utilities remains a potential surface for argument injection if inputs are not strictly validated.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8). It queries external WHOIS servers and processes the returned raw text to extract fields like 'Registrar' or 'Organization'. Maliciously crafted WHOIS records could contain instructions designed to manipulate the agent's behavior when it reads the terminal output.
Ingestion points: External WHOIS server responses captured in SKILL.md (Step 2) and saved to $OUTFILE.
Boundary markers: None. Parsed fields are echoed directly to the terminal output without delimiters or warnings to the agent to ignore embedded instructions.
Capability inventory: The skill has the capability to write files (mkdir, redirect operators), resolve DNS (dig), and query network services (whois).
Sanitization: None. The script uses xargs to trim whitespace but does not filter for terminal escape sequences or natural language instructions embedded in the data.

whois-lookup