The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the GitLab API, such as issue titles and descriptions, which serves as a surface for indirect prompt injection attacks.
Ingestion Points: Data is ingested via glab issue view, glab issue list, and glab issue note commands. Specifically, scripts/create-mr-from-issue.sh fetches the issue title from the API to generate branch names and commit messages.
Boundary Markers: The skill explicitly instructs the agent to "read and approve the file contents before posting them" and to "review the complete body and attachments... before approving the post."
Capability Inventory: The skill has the capability to execute shell commands (glab, git), write to the local file system (via redirections and scripts), and perform network operations through the glab CLI.
Sanitization: The scripts/create-mr-from-issue.sh script implements sanitization for issue titles using tr and sed to ensure they are safe for use as Git branch names, though it uses the raw title in commit messages.
[COMMAND_EXECUTION]: The skill utilizes shell scripts to automate multi-step workflows involving the glab and git command-line tools.
Evidence: scripts/batch-label-issues.sh and scripts/create-mr-from-issue.sh facilitate batch updates and branch management. These scripts use proper double-quoting for all shell variables (e.g., "$LABEL", "$ISSUE_TITLE", "$ISSUE_ID") to prevent word splitting and command injection vulnerabilities during execution.

glab-issue