The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute PowerShell scripts using pwsh -NoProfile. This allows for broad command execution capabilities on the host system.
[COMMAND_EXECUTION]: The instructions for the agent include hardcoded absolute file paths targeting a specific user's home directory (C:\Users\bobby\.claude\skills\speckit-updater\scripts\update-wrapper.ps1). This reveals local environment details and creates a dependency on a specific local path that could be leveraged for path-based attacks or indicate targeting.
[EXTERNAL_DOWNLOADS]: The skill fetches templates and update information from the GitHub Releases API at runtime.
[REMOTE_CODE_EXECUTION]: The skill is designed to download and apply updates to project files from a remote source (GitHub). If the downloaded content contains executable scripts or configuration files that the agent subsequently processes or executes, this constitutes a remote code execution vector.
[PROMPT_INJECTION]: The skill relies on parsing specific markers ([PROMPT_FOR_APPROVAL], [PROMPT_FOR_INSTALL]) from script output and external data from GitHub. This creates a surface for indirect prompt injection where malicious content from the project files or the remote API could manipulate the agent's decision-making process or trigger the -Proceed flag without valid intent.

speckit-updater