vp-macos-clean-uninstall

Warn

Audited by Gen Agent Trust Hub on May 8, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes 'sudo' to perform administrative actions such as unloading system-level launch daemons via 'launchctl' and removing package receipts via 'pkgutil'.
  • [REMOTE_CODE_EXECUTION]: The removal workflow is designed to locate and execute 'bundled uninstallers' found within third-party application bundles (e.g., '.app/Contents/'), which constitutes execution of code external to the skill.
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it incorporates untrusted filesystem data into shell commands and agent context.
  • Ingestion points: Application filenames, directory names, and bundle identifiers are ingested from the filesystem using 'find', 'mdls', and 'defaults read' in Phase 1 and Phase 3.
  • Boundary markers: The skill uses labeled output sections (e.g., '=== Homebrew cask ===') to delimit data from different detection sources.
  • Capability inventory: The agent has the ability to delete files ('rm -rf'), move files to the Trash, and execute privileged system commands ('sudo').
  • Sanitization: While the skill includes presence checks for variables (e.g., ': ${APP_NAME:?}') and uses delimiters for some commands, it does not consistently escape or sanitize filenames and bundle IDs before they are interpolated into shell strings.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 8, 2026, 11:53 AM
Security Audit — agent-trust-hub — vp-macos-clean-uninstall