security-threat-guide
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: Comprehensive review of the skill files confirms an absence of malicious instructions or security vulnerabilities. The content is strictly limited to instructional logic for threat modeling.
- [NO_CODE]: The skill consists entirely of Markdown instructions and reference documentation, with no executable scripts or code files provided.
- [PROMPT_INJECTION]: The skill instructions include explicit refusal criteria that prevent the AI from generating exploit code, security controls, or binary safety approvals, thereby reinforcing the agent's safety alignment.
- [DATA_EXFILTRATION]: No evidence of unauthorized data access or exfiltration was found. The skill maintains conversation state using a local memory file, which is a standard functional requirement for agentic tasks.
- [REMOTE_CODE_EXECUTION]: There are no patterns of remote code execution, package installation, or external resource fetching. Examples of dangerous code in the reference materials are for illustrative purposes and are not executed by the skill.
Audit Metadata