The Agent Skills Directory

[COMMAND_EXECUTION]: The Makefile templates include targets that execute shell commands with elevated privileges.
The setup target in references/gnaro_makefile.md and assets/cross_platform_guide.md uses sudo for system package management, including apt update, apt install, and a system-wide apt upgrade -y command.
The check target in references/gnaro_makefile.md executes the compiled project binary through valgrind using sudo, which grants the application root privileges during memory analysis.
[EXTERNAL_DOWNLOADS]: The skill provides templates that fetch installation and configuration scripts from external domains.
Fetches the LLVM setup script from apt.llvm.org.
References the official Homebrew installation script located on GitHub.
[REMOTE_CODE_EXECUTION]: The provided guides and reference files contain patterns for downloading and running remote scripts.
The setup target in the reference Makefile downloads llvm.sh and executes it using sudo.
The cross-platform guide for macOS includes a command that pipes a remote shell script directly from GitHub into a bash shell.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by processing and modifying external project files.
Ingestion points: The skill is intended to analyze and modify existing project Makefiles and source code as described in SKILL.md.
Boundary markers: There are no explicit instructions or delimiters provided to ensure the agent ignores potentially malicious instructions embedded in the processed user files.
Capability inventory: The skill possesses significant capabilities, including the ability to write files and execute system commands with elevated privileges via the generated Makefile targets.
Sanitization: No validation or sanitization of input files or the resulting shell commands is implemented.

modern-c-makefile