The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local git commands to manage worktrees using spawnSync with an argument array, which prevents shell injection.
Evidence: scripts/worktree.mjs executes git rev-parse, git status, and git worktree add to create isolated environments.
[INDIRECT_PROMPT_INJECTION]: The skill accepts a --prompt parameter which is passed into a new agent session. While this allows external data to influence agent behavior, it is the primary intended function of the skill.
Ingestion points: flags.prompt in scripts/worktree.mjs accepts user-supplied text.
Boundary markers: None present.
Capability inventory: File system modification via git worktree add and execution of git commands.
Sanitization: The slug parameter is sanitized via slugify(), and spawnSync is used safely, though the prompt content is passed as-is to the startSession function.

conductor-worktree