The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a local script scripts/release-workflow to execute Git commands, process manifests, and interact with the GitHub CLI (gh). Analysis of the script shows it implements several security best practices: it uses isolated temporary worktrees for modifications, performs strict validation of semantic versions, uses jq for safe JSON handling, and verifies repository-relative paths to prevent directory traversal. The script also includes a 'compare-and-swap' mechanism using input digests to ensure that the repository state hasn't changed between the user's preview and the final execution.
[SAFE]: The skill implements a robust 'approval required' model. It explicitly instructs the agent to present a diff to the user and ask for direct confirmation before pushing any code to a remote branch. The use of --dry-run modes by default further reduces the risk of accidental or unintended modifications. The script also includes comprehensive error handling and reconciliation logic to handle authentication failures or race conditions gracefully.

bumping-version