The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The skill instructs the agent to execute a local bash script scripts/worktree-manager.sh which is not included in the provided files. This makes the execution logic opaque and prevents verification of how user-supplied inputs (like branch names) are handled.
[DATA_EXFILTRATION] (LOW): The skill specifically automates the copying of .env files. Since these files typically contain sensitive API keys and credentials, their management through an unprovided and unverified script is a risk factor.
[PROMPT_INJECTION] (LOW): (Category 8 Evidence Chain) 1. Ingestion points: The skill accepts branch-name and source-branch as untrusted inputs from the user. 2. Boundary markers: None are present in the provided documentation to prevent argument injection. 3. Capability inventory: The skill has the capability to execute shell scripts via the bash command. 4. Sanitization: Sanitization cannot be verified because the execution script scripts/worktree-manager.sh is missing from the skill payload.

git-worktree