The Agent Skills Directory

[SAFE]: The skill is primarily educational, serving as a template and instruction manual for developers. It does not contain any executable scripts, external dependencies, or network-calling code within its core instructions.
[COMMAND_EXECUTION]: The skill provides examples of using shell interpolation (the ! syntax) to pre-gather context such as git diffs and logs. While this feature executes shell commands at skill load time, the examples are limited to standard, local development tools (git) and the documentation explicitly advises users to follow the principle of least privilege by restricting allowed-tools for such commands.
[PROMPT_INJECTION]: The skill defines a code review workflow that handles untrusted data (source code changes). It incorporates security best practices by explicitly instructing the agent to evaluate the code for 'hostile input' and 'trust boundary' validation, which acts as a defense against indirect prompt injection in the resulting automated reviews.

slash-commands