The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions include the installation of external tools using pip3 install sslyze and brew install testssl. Both sslyze and testssl.sh are well-known, industry-standard open-source utilities for auditing cryptographic configurations and are not considered malicious.
[COMMAND_EXECUTION]: The skill uses a variety of common networking and security tools, including curl, dig, nmap, openssl, and swaks. These tools are used for their intended purposes: performing DNS lookups, inspecting SSL/TLS handshakes, auditing HTTP headers, and testing email security protocols.
[INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting and displaying data from untrusted external sources, which presents a surface for indirect prompt injection.
Ingestion points: The skill reads DNS records (AXFR, TXT, CNAME) from target nameservers, HTTP response headers from target domains, and certificate data from certificate transparency logs (via crt.sh).
Boundary markers: The instructions do not implement specific delimiters or 'ignore instructions' warnings to prevent the agent from being influenced by data content parsed from these sources.
Capability inventory: The skill environment includes the ability to execute shell commands, perform network requests, and install system/Python packages.
Sanitization: There is no evidence of data sanitization or validation performed on the external content before it is presented to the agent.
[REMOTE_CODE_EXECUTION]: An automated security scan flagged the pattern of piping curl output to python3 as a remote code execution risk. However, manual inspection of the skill's source code reveals that these pipes are directed specifically to python3 -m json.tool. This invocation uses a standard Python utility module to format JSON data for readability and does not execute the input stream as code.

hunt-tls-network