ad-test-designer

Pass

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies and mitigates potential risks from untrusted data ingestion.
  • Ingestion points: Processes user-supplied CSV exports for experiment results (SKILL.md).
  • Boundary markers: Explicit instructions are provided to the agent to treat CSV content strictly as data values and ignore any potential embedded commands.
  • Capability inventory: The skill can execute a local Python script for calculations and write to a 'memory/' directory with user consent.
  • Sanitization: Employs clear instructional boundaries to separate data from commands, minimizing the risk of indirect injection.
  • [COMMAND_EXECUTION]: The skill invokes a local statistical helper script (experiment.py) with specific, non-arbitrary arguments. There is no provision for executing user-defined shell commands or external binaries.
  • [REMOTE_CODE_EXECUTION]: No remote code execution or external downloads were detected. All scripts used are local to the plugin environment.
  • [DATA_EXFILTRATION]: The skill does not contain any network-facing code or patterns that suggest unauthorized data transmission. Data processing is kept within the agent's local execution context.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 20, 2026, 07:29 PM
Security Audit — agent-trust-hub — ad-test-designer