The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by reading configuration preferences from local files.
Ingestion points: The skill reads worktree directory preferences from CLAUDE.md using grep.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present when reading from external project files.
Capability inventory: The skill has access to execute git, npm, cargo, pip, poetry, and go commands for environment setup and testing.
Sanitization: There is no evidence of sanitization or validation of the configuration strings retrieved from CLAUDE.md before they are used in path construction.
[COMMAND_EXECUTION]: The skill executes standard development toolchains (npm, cargo, pip, poetry, go) based on the detection of project manifest files. This is consistent with the skill's primary purpose of initializing isolated development environments.

using-git-worktrees