The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs standard grep commands to perform local discovery of project metadata (e.g., target framework, C# version) and to scan source code for performance-related patterns. These operations are limited to the local filesystem and do not involve privilege escalation or dangerous flags.
[DATA_EXPOSURE]: No sensitive file access or network-based data exfiltration patterns were found. The skill only processes project configuration and source code for the purpose of generating an optimization report.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute remote scripts. While it recommends several external C# libraries (e.g., UniTask, MemoryPack, ZLinq), these are established open-source performance tools recommended for the user to integrate into their own project, rather than being executed by the skill itself.
[PROMPT_INJECTION]: Analysis of the skill instructions and reference materials revealed no attempts to override agent safety protocols, extract system prompts, or bypass behavioral constraints.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by reading the user's codebase. However, its capabilities are restricted to identifying text patterns and generating a report based on those findings. It does not execute the code it reads, minimizing the risk of indirect instruction execution.

csharp-perf-optimizer