The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/worktree.py reads shell commands from a .worktreerc.yml file in the repository root and executes them using subprocess.run(shell=True). This allows for arbitrary shell command execution based on local configuration files.
[REMOTE_CODE_EXECUTION]: The skill enables execution of commands defined within the project being worked on. If a repository contains malicious instructions in its configuration file, the agent will execute them in the user's environment.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to instructions embedded in the repository configuration that influence the agent's actions.
Ingestion points: scripts/worktree.py reads .worktreerc.yml (or .yaml) via the load_worktreerc function.
Boundary markers: Absent. The script does not validate or sanitize the commands before execution.
Capability inventory: scripts/worktree.py uses subprocess.run(shell=True) for command execution and shutil.copytree/shutil.copy2 for file system operations.
Sanitization: Absent. The commands are passed directly to the shell as provided in the YAML file.
[DYNAMIC_EXECUTION]: The skill dynamically executes shell commands at runtime based on the contents of the post_create section in the project's YAML configuration file.

wt-create