gjkim-instruction

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: No override markers or jailbreak attempts were detected. The instructions are focused on maintaining a structured markdown document named gjkim_instruction.md.
  • [DATA_EXFILTRATION]: The skill does not perform network operations or access sensitive system directories. It interacts only with a specific project document.
  • [REMOTE_CODE_EXECUTION]: There are no external dependencies, package installations, or remote script execution patterns in the skill.
  • [COMMAND_EXECUTION]: No shell commands, privilege escalation (sudo), or persistence mechanisms (cron/bashrc) are present.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user input to populate the goal and requirements sections of the document. While this is a data ingestion surface, the skill implements a strict template and limits content to markdown text, posing no risk of automated execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 06:00 AM
Security Audit — agent-trust-hub — gjkim-instruction