magpie-setup-isolated-setup-install

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • Sandbox Control Bypass: The skill instructions direct the agent to intentionally bypass default security boundaries for specific file operations.
  • Evidence: The text explicitly instructs the agent to use the dangerouslyDisableSandbox: true flag when writing to project-local settings files to resolve platform-specific directory access limitations.
  • Indirect Prompt Injection Surface: The skill is designed to ingest and follow instructions from an external overrides file located within the adopter repository.
  • Ingestion points: .apache-magpie-overrides/setup-isolated-setup-install.md (relative to the repository root).
  • Boundary markers: No specific delimiters or warnings are defined to separate external configuration instructions from the skill's primary logic.
  • Capability inventory: The skill possesses the capability to modify environment variables, change shell configurations, and execute shell commands via the Bash tool.
  • Sanitization: There is no mention of validation or sanitization procedures for the contents of the overrides file before the agent processes them.
  • Persistence and System Configuration Modification: The skill guides the user through making permanent changes to the host environment.
  • Evidence: The walkthrough includes steps to modify shell initialization files (such as ~/.bashrc or ~/.zshrc) and configure global Git hook paths using git config --global core.hooksPath.
  • Local Environment Scanning: The skill performs discovery of local project structures to automate configuration across multiple repositories.
  • Evidence: It uses directory traversal with find and git worktree list to identify existing Git checkouts across common user directories.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 12:15 PM
Security Audit — agent-trust-hub — magpie-setup-isolated-setup-install