benchmark-prd-scout

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is designed for structured product analysis and benchmarking tasks as described in its metadata and documentation.
  • [PROMPT_INJECTION]: The skill implements a defensive activation policy requiring explicit user intent (e.g., using '$benchmark-prd-scout'). Furthermore, Section 5 of the 'Operating Principles' provides clear instructions to treat external research data (READMEs, issues, websites) as inert data, preventing potential override instructions in those sources from hijacking the agent's behavior.
  • [COMMAND_EXECUTION]: No shell commands, system calls, or unauthorized script execution patterns were identified within the skill's instructions or configuration files.
  • [DATA_EXFILTRATION]: There are no indicators of unauthorized data collection. Network operations are confined to legitimate research via standard browser tools for analyzing URLs provided by the user.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 04:12 AM
Security Audit — agent-trust-hub — benchmark-prd-scout