The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The skill instructs the agent to run ./format.sh, an opaque local script within the project directory. This represents a risk of arbitrary command execution as the script's contents are not defined or verified by the skill.
[COMMAND_EXECUTION] (HIGH): The documentation encourages the use of sudo apt-get install, which is a privilege escalation risk. Granting root access to install software can be exploited if the package name or source is manipulated.
[EXTERNAL_DOWNLOADS] (LOW): Recommends downloading tools via apt-get and brew. While these are generally trusted sources, the action of downloading and installing software during skill execution should be handled with caution.
[INDIRECT_PROMPT_INJECTION] (HIGH): The skill processes external C++ source files and project configuration files with the capability to modify the filesystem and execute scripts.
Ingestion points: C++ source files located in src/, include/, and tests/ and configuration files like .clang-format.
Boundary markers: Absent; there are no instructions to the agent to disregard malicious instructions embedded in code comments or configuration values.
Capability inventory: Includes file system modification (clang-format -i), local script execution (./format.sh), and Git operations.
Sanitization: Absent; the skill does not validate the integrity or safety of the code files before processing them with executable tools.

C++ Code Formatter