prompt-variation-guide
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides structured templates and frameworks designed to be populated with external input (e.g., placeholders like
[직업명],[작업명], and[제약조건]inreferences/techniques.md). This interpolation of data into prompt structures creates a surface for indirect prompt injection. - Ingestion points: Data placeholders and "Universal Framework" layers in
SKILL.mdandreferences/techniques.md. - Boundary markers: The skill suggests using XML-like tags (e.g.,
<system_prompt>,<constraints>,<instructions>) as structural delimiters, though they are not explicitly defined as security boundaries to prevent instruction override. - Capability inventory: No executable code, shell commands, file writing, or network operations are present in any of the skill's files.
- Sanitization: No instructions for input validation, escaping, or filtering of user-provided content are included in the documentation.
- [NO_CODE]: The skill consists entirely of Markdown files providing guidance and templates, with no executable scripts, binaries, or automated tool invocations.
Audit Metadata