The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted C/C++ source code provided by the user, which could theoretically contain embedded instructions designed to confuse the analysis scripts.
Ingestion points: Source code files are read by scripts/circular_header_check.py and scripts/scan_cpp_size.py using Path.read_text().
Boundary markers: The analysis scripts do not implement specific boundary markers or instructions to ignore embedded AI prompts within the source code comments.
Capability inventory: The skill's capabilities are restricted to file system read access for analysis and optional file system write access for report generation. It contains no network access, shell execution, or privilege escalation capabilities.
Sanitization: The scripts use regular expressions for static analysis and do not perform content sanitization, but the lack of executable capabilities makes this risk negligible.

code-checker