skill-manager

Warn

Audited by Gen Agent Trust Hub on Oct 15, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPERSISTENCEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PERSISTENCE]: The skill maintains persistence on macOS by registering a LaunchAgent (~/Library/LaunchAgents/io.testany.skilldock.update...plist). This mechanism is used to run background tasks that check for and apply updates to installed skills and plugins according to a user-defined schedule.
  • [COMMAND_EXECUTION]: SkillDock spawns multiple subprocesses to perform management tasks. It executes the git binary for repository operations, the codex CLI for OpenAI plugin management, and the claude CLI for Anthropic plugin operations. It also executes Node.js binaries found on the system to run its own server and build processes.
  • [EXTERNAL_DOWNLOADS]: The application fetches updates and marketplace data from external sources, primarily GitHub and the npm registry. It also includes logic to download and verify Node.js runtimes if a compatible version is not found locally. These downloads are directed toward official or user-specified repositories.
  • [INDIRECT_PROMPT_INJECTION]: The tool ingests and parses SKILL.md files and plugin metadata from various local and remote sources to display information in the GUI. While this creates an exposure to potentially malicious instructions embedded in third-party content, the tool treats these as data for display and does not execute the content of the skills it manages.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 15, 2026, 08:27 PM