worktree
Warn
Audited by Gen Agent Trust Hub on Mar 15, 2026
Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill dynamically constructs and executes shell commands using variables derived from user input ($NAME and $N).
- Evidence: The script constructs file paths and branch names using string concatenation:
WT_PATH="$PARENT_DIR/$REPO_NAME.wt-$NEXT-$NAME"andBRANCH="agents/$NEXT-$NAME". These are subsequently passed togit worktree addandgit worktree remove. - Risk: If the AI agent does not sanitize the user-provided arguments, an attacker could inject shell metacharacters (e.g.,
;,$(),`) to execute arbitrary commands. Additionally, the use of../in the name field could allow path traversal, enabling the creation or deletion of directories outside the intended parent directory.
Audit Metadata