bestie-features

Pass

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads and executes the 'bestie-template' package from a remote Python registry using the 'uvx' tool to provide its core functionality.
  • [COMMAND_EXECUTION]: The skill uses shell commands to interact with the system environment, including 'git' for status and diffing, and the 'bestie' CLI for applying package templates.
  • [PROMPT_INJECTION]: The skill processes untrusted metadata from local files such as 'Project.toml' and 'git remote' output to populate arguments for CLI commands. This represents an indirect prompt injection surface.
  • Ingestion points: Extracts data from 'Project.toml', '.copier-answers.yml', and 'git remote -v' outputs.
  • Boundary markers: No specific delimiters are used to wrap the untrusted data when passed to shell commands.
  • Capability inventory: Full shell execution capability via the 'uvx' wrapper.
  • Sanitization: The instructions mandate that the agent must confirm guessed or extracted values with the user before applying changes, providing a manual verification step.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 8, 2026, 02:37 PM
Security Audit — agent-trust-hub — bestie-features