The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions utilize sudo tcpdump for network traffic capture, which requires the agent to operate with elevated privileges. While standard for network analysis, it increases the risk of privilege abuse if the environment is compromised.
[EXTERNAL_DOWNLOADS]: The skill directs the user to download a third-party Ghidra extension (QtREAnalyzer) from a personal GitHub repository and to install the Kaitai Struct compiler via yay (Arch User Repository). These sources are not officially verified and could introduce malicious code into the analysis environment.
[REMOTE_CODE_EXECUTION]: The workflow involves downloading and running external tools and scripts (such as Ghidra and the QtREAnalyzer plugin) to analyze potentially malicious closed-source binaries. Executing code from unverified sources to handle untrusted input presents a risk of remote code execution.
[PROMPT_INJECTION]: The skill is designed to process and analyze untrusted data from binary files and network captures, creating an attack surface for indirect prompt injection. \n * Ingestion points: Binary file headers and strings in scripts/analyze-binary-format.sh, Ghidra project imports, and raw network packet captures. \n * Boundary markers: The skill lacks defined delimiters or explicit 'ignore' instructions to isolate binary data from being interpreted as instructions by the agent. \n * Capability inventory: The skill has access to shell execution (Bash), file system modification (Read, Write, Edit), and interaction with specialized debuggers. \n * Sanitization: No specific validation or filtering of binary content is performed before the agent processes the findings.

code-re-qt5