The Agent Skills Directory

COMMAND_EXECUTION (HIGH): The skill instructs the agent to perform sensitive shell operations, including recursive directory deletion (rm -rf) and running local project scripts (npm run dev). If these commands are executed on a repository containing malicious configurations or if the agent misinterprets file paths, it can lead to data loss or arbitrary code execution.
EXTERNAL_DOWNLOADS (HIGH): The PR review workflow includes npm install, which downloads and executes third-party packages from the NPM registry. Since the skill is designed to work with external branches (e.g., origin/feature-y), it implicitly trusts code from external sources, making it vulnerable to supply chain attacks.
REMOTE_CODE_EXECUTION (HIGH): By combining git fetch (fetching remote code) with npm install and npm run dev (executing that code), the skill creates a direct pipeline for Remote Code Execution. An attacker could submit a PR with a malicious postinstall script or a poisoned dev script that executes when the agent follows the skill's workflow instructions.

git-worktree