The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied C++ source code and pull request details, which is an inherent attack surface for indirect prompt injection through comments or metadata. * Ingestion points: User-provided C++ source files and PR descriptions (referenced in SKILL.md). * Boundary markers: No explicit boundary markers or safety warnings for embedded instructions are provided in the skill instructions. * Capability inventory: The skill utilizes subprocess execution for standard developer tools including compilers, debuggers, and profilers (found in references/tooling-build.md and references/concurrency.md). * Sanitization: The instructions do not define sanitization procedures for the external code being analyzed.
[COMMAND_EXECUTION]: The skill provides command templates for standard system diagnostic and binary inspection tools. * Evidence: The skill suggests the use of tools like objdump, nm, readelf, strace, and valgrind for legitimate debugging, profiling, and binary analysis tasks.

cpp-patterns