setup

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines an experiment setup workflow where user-supplied parameters—such as domain, name, and evaluation commands—are interpolated into a shell command string for execution. This creates a potential surface for indirect prompt injection or command injection if inputs are not strictly validated by the agent platform.\n
  • Ingestion points: User-provided values for domain, name, target file path, and evaluation command collected via the /ar:setup interactive prompt in SKILL.md.\n
  • Boundary markers: No delimiters or specific safety instructions for validating these inputs are included in the skill body.\n
  • Capability inventory: Shell command execution using a local Python script setup_experiment.py defined in the skill's scripts directory.\n
  • Sanitization: No explicit sanitization or filtering logic is specified for the input parameters before they are used in the command template.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 02:34 PM
Security Audit — agent-trust-hub — setup