The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for the user to install security tools including httpx, subjack, and nuclei via go install from their respective GitHub repositories. These tools are industry standards for reconnaissance and originate from well-known security organizations like ProjectDiscovery.
[COMMAND_EXECUTION]: The skill executes network-oriented commands such as dig and curl. It also constructs file paths using the SECSKILL_TARGET_DOMAIN environment variable (e.g., "$OUTDIR/cnames_${TARGET}.txt"). The absence of sanitization on this variable could permit path traversal if it contains directory navigation characters, potentially allowing the skill to write files outside of the intended output directory.
[PROMPT_INJECTION]: The skill processes external data from a subdomain list (SECSKILL_SUBDOMAIN_LIST) and incorporates it into shell commands and automated scan logic. This represents a potential surface for indirect prompt injection if the source data is malicious.
Ingestion points: The subdomain list file is read in Steps 2, 3, 4, and 5 to provide targets for DNS lookups, HTTP probes, and automated scanners.
Boundary markers: No delimiters or boundary markers are used to separate untrusted data from the command context.
Capability inventory: The skill has the ability to perform DNS resolution, initiate HTTP/HTTPS requests, and execute external binary tools.
Sanitization: There is no evidence of sanitization or validation of the input subdomain strings before their use in command execution.

subdomain-takeover