style-transfer
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text provided by the user in the
textandstyle_referencefields, which serves as a potential surface for indirect prompt injection attacks. - Ingestion points: External data is processed via the
textandstyle_referenceinputs inSKILL.md. - Boundary markers: There are no explicit instructions to the agent to treat input text as data only or to ignore embedded instructions (delimiters are absent).
- Capability inventory: The skill uses the file system to write results into the
outputs/directory. - Sanitization: The instructions do not define any sanitization, filtering, or validation steps for the processed input content.
- [EXTERNAL_DOWNLOADS]: The
EASEL-META.mdfile contains several URLs to public GitHub repositories used for academic reference and research attribution. - References: The skill documents links to
github.com/comp-int-hum/llm-style-transfer,github.com/fuzhenxin/Style-Transfer-in-Text, andgithub.com/souro/tst_llmas foundational sources for its logic.
Audit Metadata