rails-upgrade-7-to-8

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines instructions for the agent to generate shell commands, such as using sed to update version pins in project files, based on user-supplied information regarding their Rails environment.
  • Ingestion points: User prompts identifying their current Rails version and migration targets (referenced in SKILL.md metadata and usage instructions).
  • Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' instructions to isolate user-provided version data from the command generation logic.
  • Capability inventory: The skill patterns involve shell command execution for dependency management (bundle install), file manipulation (sed), and application testing (bin/rspec, bin/rails) across SKILL.md Patterns 1 through 6.
  • Sanitization: The skill lacks explicit validation or escaping of user-provided version strings before they are interpolated into the suggested command templates.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 03:48 AM
Security Audit — agent-trust-hub — rails-upgrade-7-to-8