The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local git commands to retrieve repository diffs. The instructions specifically mandate the use of argv lists rather than shell strings, which follows security best practices to prevent command injection vulnerabilities.
[DATA_EXFILTRATION]: Source code diffs are retrieved and transmitted to the check_changes tool for analysis. This is the intended primary function of the skill and occurs through the established MCP (Model Context Protocol) transport.
[SAFE]: The skill includes comprehensive guidelines in mcp-rules.md that explicitly forbid the agent from probing for tokens, bypassing IDE security controls, or inventing organizational policies, which significantly reduces the risk of autonomous misbehavior.

check-changes