The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Bash(git:*) and Bash(bash:*) tools to read repository history and run a local analysis script (scripts/smart-rebase-analyze.sh). These operations are scoped to read-only history inspection (git log, git rev-parse, git merge-base).
[COMMAND_EXECUTION]: The skill explicitly refuses to execute git rebase or git push directly. It instead generates a shell-quoted command string for the user to execute manually. This is documented as a security boundary to ensure human-in-the-loop verification for destructive operations.
[SAFE]: The skill implements extensive 'ref-name hardening' to prevent command injection via malicious branch names. It uses --end-of-options and shell-quoting for all git operands and validates that ref names are valid UTF-8 to prevent JSON corruption or shell misinterpretation.
[SAFE]: The skill performs mandatory prerequisite checks before generating any plan, such as ensuring the working tree is clean and verifying the user is not on a protected branch (e.g., main, master, develop).

smart-rebase