The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute system-level commands using the Windows Package Manager (winget) within a PowerShell environment.
Evidence: The skill uses winget upgrade --disable-interactivity to fetch data and winget upgrade --id "<ID>" to perform installations.
[PRIVILEGE_ESCALATION]: The skill performs actions that frequently require elevated permissions on Windows systems, which is handled via standard operating system prompts.
Evidence: The instructions explicitly state that installers may raise Windows User Account Control (UAC) prompts that the user must approve.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data generated by the winget tool, which includes package names, IDs, and versions that could theoretically be manipulated by a malicious package repository.
Ingestion points: The output of the winget upgrade command is parsed to create a numbered list for the user (SKILL.md).
Boundary markers: None present to distinguish system output from potential embedded instructions.
Capability inventory: The skill has the capability to execute PowerShell commands based on the parsed package IDs.
Sanitization: The instructions use double quotes around variables (e.g., "<ID>") when constructing commands, which provides a basic level of shell argument encapsulation, though it does not provide full sanitization against sophisticated injection in the package metadata.

winget-upgrade