The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes several shell and Python scripts (such as mr-review-workflow.sh and post-inline-comment.py) that invoke the glab CLI and local test commands. These operations are protected by input validation and command whitelists to prevent arbitrary execution.
[DATA_EXFILTRATION]: Scripts like add-inline-comment.sh and post-inline-comment.py access GitLab authentication tokens to communicate with the GitLab API. This communication is restricted to HTTPS and targets the user-configured GitLab host (defaulting to gitlab.com).
[PROMPT_INJECTION]: Documentation in glab-ci/SKILL.md and SECURITY.md identifies the risk of indirect prompt injection from untrusted GitLab data (e.g., CI logs, issue titles) and provides explicit safety instructions for agents processing this data.

gitlab-cli-guide