The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/setup_worktree.py executes shell commands defined in the setupCommand configuration field using subprocess.run with shell=True. This allows the execution of arbitrary commands in the worktree environment.
[DYNAMIC_EXECUTION]: Workspace setup involves executing commands and logic derived from project-specific configuration files (.agent-work/workspace.json). Malicious modification of these files could lead to unauthorized code execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repository data. Ingestion points: Reads ticket.md, task.md, and outline Markdown files in the task directory (SKILL.md), as well as legacy scripts/create_worktree.sh. Boundary markers: None present; instructions do not specify using delimiters for these external files. Capability inventory: The skill can execute shell commands (setupCommand), write files (copyGlobs), and perform Git operations (scripts/setup_worktree.py). Sanitization: Script logic validates copy paths, but no sanitization is performed on natural language content in read files.
[DATA_EXFILTRATION]: The skill's default behavior is to copy sensitive environment files (e.g., .env, .env.local) to new worktree locations. While intended for functional parity, this automates the proliferation of sensitive credentials.

setup-worktree