skills/ulpi-io/skills/build-dmg/Gen Agent Trust Hub

build-dmg

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes project-level files such as VERSION, Info.plist, and ExportOptions.plist. While these are standard files for the intended task, their content is interpolated into shell commands and AppleScript, which presents a surface for injection if the project being built is malicious.
  • Ingestion points: SKILL.md (Step 0) and helpers/build-dmg.sh (reading VERSION_FILE, Info.plist).
  • Capability inventory: Bash tool executes the helper; helper executes xcodebuild, hdiutil, osascript, and PlistBuddy.
  • Boundary markers: The skill instructs the agent to ask the user for confirmation if inputs are ambiguous.
  • Sanitization: The script performs minimal sanitization (whitespace stripping) of file-derived variables before interpolation into commands.
  • [DYNAMIC_EXECUTION]: The helper script helpers/build-dmg.sh dynamically generates an AppleScript using a here-doc and executes it via osascript to style the DMG window. The script interpolates the APP_NAME variable into the AppleScript body.
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to run a local helper script which in turn executes several macOS system utilities (xcodebuild, hdiutil, PlistBuddy, osascript) to automate the build and packaging process. These operations are within the stated scope of a DMG builder.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 12:23 PM
Security Audit — agent-trust-hub — build-dmg