The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines shell command templates for linting and formatting C++ code using local tools. Evidence: SKILL.md and references/tooling/toolchain.md specify the use of clang-tidy and clang-format via bash -c pipelines. Security Analysis: The commands utilize find ... -print0 and xargs -0, which is the recommended secure method for handling arbitrary filenames in shell scripts, preventing command injection via malicious file names.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process C++ source files which could contain malicious instructions. 1. Ingestion points: Local .cpp, .hpp, and .cc files discovered in the project workspace. 2. Boundary markers: The skill relies on the agent to correctly substitute the {file} placeholder with validated paths and provides a structured output template. 3. Capability inventory: File read access and execution of local binary tools (clang-tidy, clang-format) via subprocesses. 4. Sanitization: The use of null-terminated strings (-print0 and -0) in the shell pipeline serves as a technical control against injection during the file discovery and batch processing phase.

codeguard-cpp