The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs Git and GitHub CLI operations to manage repository state. Security is prioritized through explicit instructions to redact credentials from all outputs, avoid dumping sensitive environment variables or configuration files, and warn users before including potentially sensitive files (like .env, *.pem, or *.key) in commits.
[DYNAMIC_EXECUTION]: The skill executes project-specific commands for testing and linting (e.g., npm test, pytest, make test) defined in a .gitx.json configuration file or detected in the repository. It mitigates risk by requiring the agent to inspect these commands for high-risk behaviors—such as network exfiltration or file deletion—and to stop execution if the commands exceed the authorized scope.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources including GitHub issue titles, comments, and Git diffs. It includes robust defensive instructions that require the agent to treat this data solely as "untrusted reference material" and to never obey instructions embedded within that data that conflict with safety protocols or user intent.

gitx