write-judge-prompt
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFENO_CODEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [SAFE]: The skill follows established best practices for designing LLM evaluators and does not exhibit any malicious behavior, obfuscation, or safety bypass attempts.
- [NO_CODE]: The skill consists entirely of instructional markdown and configuration metadata; it contains no scripts or binary executables, which minimizes its technical attack surface.
- [EXTERNAL_DOWNLOADS]: The documentation includes an installation command for the skill itself using the author's namespace (m1nga/skills@write-judge-prompt), representing a standard and legitimate vendor-owned resource.
- [PROMPT_INJECTION]: The skill ingests user-provided traces and failure modes to generate prompts, creating a surface for indirect prompt injection. However, since the skill possesses no execution capabilities, the risk is negligible. * Ingestion points: User-provided failure mode descriptions and trace examples (SKILL.md). * Boundary markers: The skill uses standard markdown headers but does not define specific delimiters for untrusted data interpolation. * Capability inventory: No shell execution, network operations, or file-writing scripts are present. * Sanitization: The skill does not provide instructions for sanitizing or escaping the user-provided data within the generated prompts.
Audit Metadata