skills/davidondrej/skills/bb-plugins/Gen Agent Trust Hub

bb-plugins

Pass

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill documents the capability to install and execute third-party plugins using the bb plugin install command. It explicitly states that plugins are 'full-trust' and 'run in-process'. To mitigate risks, the documentation warns users to review source code before installation and notes that the --yes flag can be used to bypass security prompts.
  • [EXTERNAL_DOWNLOADS]: The skill provides examples for fetching plugins from external sources such as NPM and GitHub (e.g., https://github.com/acme/bb-plugin-notes.git). These references are instructional and target well-known services.
  • [COMMAND_EXECUTION]: The skill contains various shell commands for plugin lifecycle management, including bb plugin new, bb plugin dev, bb plugin reload, and bb plugin logs. These are standard developer tools for the associated platform.
  • [PROMPT_INJECTION]: The skill describes a surface for indirect prompt injection where plugins can contribute custom agent tools, instruction folders, and providers. The documentation provides a security guideline to review third-party source code and store sensitive credentials using specific 'secret' settings to minimize exposure.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 20, 2026, 09:25 PM
Security Audit — agent-trust-hub — bb-plugins