The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the tshark command-line utility extensively through both shell instructions in SKILL.md and a wrapper script scripts/agent.py. The Python script utilizes subprocess.run to execute commands. It follows security best practices by passing arguments as a list rather than a single string and uses shlex.split to handle complex argument strings, which mitigates simple shell injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external, untrusted data in the form of network packet captures (PCAP/PCAPNG files). A sophisticated attacker could attempt to embed malicious instructions within network packet fields (such as User-Agent strings, DNS query names, or HTTP headers) that might influence an AI agent's behavior when it processes the extracted results.
Ingestion points: scripts/agent.py reads data from PCAP files via tshark -r and extracts various protocol fields.
Boundary markers: The current implementation does not explicitly use delimiters or "ignore embedded instructions" warnings when processing extracted packet content.
Capability inventory: The skill possesses the ability to execute system commands (tshark), write files to the local directory (e.g., extracted_file_hashes.txt, urls.txt), and perform network captures.
Sanitization: The Python implementation uses shlex.split for command argument preparation, but the content extracted from the packets is split by delimiters (e.g., |) without further validation before being returned as structured data.

analyzing-network-traffic-with-wireshark