The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to process untrusted data (malware samples, PCAP logs) which could contain indirect prompt injections (Category 8).
Ingestion points: The skill facilitates reading malware binaries, scripts, and network captures through analysis tools and scripts (e.g., as described in references/scripts-and-obfuscation.md).
Boundary markers: No explicit delimiters or system instructions are provided to the agent to ignore potentially malicious instructions embedded within the analyzed samples.
Capability inventory: The skill utilizes powerful system capabilities including shell command execution (tshark, vol3, peframe), filesystem access, and the ability to run external binaries with elevated privileges (sudo tcpdump).
Sanitization: There is no explicit sanitization or validation of the data extracted from analyzed samples before it is processed by the agent or presented to the user.
[COMMAND_EXECUTION]: The skill includes instructions and Python code snippets for executing and monitoring potentially malicious files using tools like strace, ltrace, and sudo tcpdump. While these are high-risk operations, they are directly related to the skill's primary purpose of malware analysis.
[EXTERNAL_DOWNLOADS]: The skill references external, third-party tools hosted on GitHub (e.g., PyArmor-Unpacker) for specific malware analysis tasks.

ctf-malware