prompt-master
Pass
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill defines an indirect prompt injection surface by design.
- Ingestion points: Untrusted data enters via user-provided prompts, contexts, and constraints as seen in the workflow of SKILL.md.
- Boundary markers: Mitigation is attempted using XML tagging and instructions to keep directives outside user-data tags.
- Capability inventory: No dangerous tools or code capabilities (e.g., shell access, file writes) are utilized in this skill.
- Sanitization: No input sanitization is present. Recommended practice includes using strict delimiters and instructing the agent to ignore any embedded directives.
- [NO_CODE]: The skill contains only instructional markdown and templates, with no scripts or executable logic, which minimizes the potential for technical exploitation.
Audit Metadata