The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository environment, such as branch names and test command configurations. If these inputs contain shell metacharacters, they could result in command injection when interpolated into the provided bash templates.
Ingestion points: Branch names are identified and used in Step 3 and Step 5, and repository-specific test commands are identified and executed in Step 1.
Boundary markers: The skill uses backticks for code blocks but does not provide explicit instructions to sanitize or escape variables used within these bash scripts.
Capability inventory: The skill utilizes bash for environment detection, merging, and workspace cleanup operations, including destructive commands like git worktree remove and git branch -d.
Sanitization: No sanitization logic or validation is specified for external inputs before their use in shell commands.
[COMMAND_EXECUTION]: The skill defines several bash scripts intended for the agent to execute. These scripts interact with the local git environment and file system to manage worktrees and branches. While these operations are core to the skill's purpose, they represent a significant capability tier that operates on environment variables and user-selected options.

finishing-a-development-branch