commerce-marketing-compliance-kr

Pass

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill includes instructions that attempt to influence platform-level execution and processes untrusted user data without sufficient isolation.
  • Bypass Directive: The frontmatter description includes the directive 'ai-slop-reviewer 체이닝 제외' (Exclude from ai-slop-reviewer chain), which attempts to bypass a platform quality or safety review mechanism.
  • Indirect Prompt Injection Surface: The skill is designed to ingest and process user-supplied content ('Message Drafts') which can contain malicious instructions intended to manipulate the compliance verdict.
  • Ingestion points: Defined in SKILL.md under the '입력 슬롯' (Input Slots) section as '메시지 초안' (Message Draft).
  • Boundary markers: Absent; there are no instructions for the agent to use delimiters like triple quotes, XML tags, or explicit 'ignore instructions' warnings when handling the draft input.
  • Capability inventory: No executable tools, network capabilities, or file system access are enabled for this skill.
  • Sanitization: No validation or escaping of the user-provided message draft is performed prior to analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 22, 2026, 06:27 AM
Security Audit — agent-trust-hub — commerce-marketing-compliance-kr