The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected in the skill instructions. The skill performs benign architectural reviews and documentation tasks.
[COMMAND_EXECUTION]: The skill uses slash commands with $ARGUMENTS to specify targets for review (e.g., /audit-architecture $ARGUMENTS). This is a standard mechanism for user-provided input in development workflows and does not present an injection risk given the skill's restricted output actions.
[DATA_EXPOSURE]: The skill reads standard project files like README, CLAUDE.md, and directory structures to perform analysis. It does not attempt to access sensitive directories such as .ssh or .aws, nor does it perform any network exfiltration operations.
[NO_CODE]: The skill instructions explicitly state that it does not write or implement code, focusing entirely on analysis and documentation, which significantly reduces its attack surface.

bezaleel-architect