The Agent Skills Directory

[SAFE]: The skill performs static analysis of code and generates reports. It explicitly mandates that the code reference is read-only and forbids modifications to the working tree, git state, or execution of tests.
[SAFE]: No network operations, external data exfiltration, or remote downloads are identified. The skill does not define any external dependencies or package installations.
[SAFE]: The skill does not request administrative privileges (sudo) or attempt to establish persistence on the host system.
[SAFE]: There are no signs of prompt injection, obfuscation, or malicious dynamic context injection using shell placeholders.
[SAFE]: While the skill processes external code which could theoretically contain indirect prompt injection, this is the core intended functionality of a code review tool and the skill's capabilities are limited to writing local markdown reports.

review-code