The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool for project analysis. This access is restricted to searching and scanning code for performance patterns (using grep, glob, etc.) and does not involve executing untrusted shell scripts or performing network-based command injection.
[DATA_EXFILTRATION]: The skill accesses configuration and reference files located in a parent directory (../rc-codeprobe/). This is a legitimate use of file system access to load shared constraints and tech-stack-specific guidelines required for the performance audit.
[PROMPT_INJECTION]: As the skill is designed to ingest and analyze untrusted source code, it possesses an inherent attack surface for indirect prompt injection. However, its instructions are focused on structured detection of performance anti-patterns, which significantly mitigates the risk of the agent obeying instructions embedded in analyzed data.

rc-codeprobe-performance