The Agent Skills Directory

[COMMAND_EXECUTION]: The skill mentions the Bash tool in the allowed-tools section and references a Research Protocol that uses Grep and Glob. These are used for static analysis of existing local source code to ground the architectural review, which is an expected and safe behavior for this task.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided descriptions of architectural changes ($ARGUMENTS). While this is a data ingestion point, the skill's instructions strictly focus on analytical frameworks (SWOT, STRIDE, C4, etc.) and reading local source files. There are no capabilities for network exfiltration or modifying sensitive system files that could be exploited by an injection attack.

architecture-review