refactor-safety

Pass

Audited by Gen Agent Trust Hub on Apr 21, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is a documentation-based safety guide for refactoring. It does not contain executable malicious code, obfuscation, or data exfiltration patterns.
  • [COMMAND_EXECUTION]: The skill references file system operations like Read and Grep to inspect source code (e.g., src/views/Order.vue). These are standard for coding-related AI skills and are used here for safety verification, not for accessing sensitive system files.
  • [DATA_EXFILTRATION]: There are no network-related commands or instructions to send data to external servers.
  • [PROMPT_INJECTION]: The instructions reinforce existing safety behaviors by requiring the agent to verify its work against original code, rather than attempting to bypass safety filters.
  • [SAFE]: Indirect prompt injection (Category 8) risk is minimal; while the skill reads source files, the capabilities are limited to standard code analysis tasks without network access or high-privilege execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 21, 2026, 12:57 AM
Security Audit — agent-trust-hub — refactor-safety