gardener-tasks

Pass

Audited by Gen Agent Trust Hub on Nov 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • Indirect Prompt Injection: The agent is designed to process content from GitHub issues, pull requests, and comments, which can be provided by any user. This represents a potential surface for indirect prompt injection. The skill includes specific instructions to treat all external text as untrusted data and to disregard any instructions or commands embedded within it.
  • Command Execution via Shell: Multiple tasks utilize the repository.exec tool, allowing the agent to run shell commands to build projects, run tests, and format code. While this is a common requirement for a developer-oriented agent, it grants the skill significant control over the execution environment.
  • External Repository Cloning and Execution: Tasks such as pi-upgrade and opencode-upgrade are configured to clone external GitHub repositories to identify and test dependency updates. This process involves executing build and test commands that may run code from those external sources, which is necessary for the skill's core upgrade functionality.
  • Dynamic Package Management: The skill dynamically installs specific versions of tools like Node.js and pnpm using npx during its execution. This involves fetching software from public registries to ensure the environment matches the required configuration.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 7, 2026, 10:23 PM