one-sentence-test
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill is entirely instructional and does not contain any executable scripts, command-line operations, or network requests. It serves as a mental model for the AI to follow during design reviews.
- [PROMPT_INJECTION]: The instructions include phrases such as "Ignore the name, JSDoc, README" and "Don't skip." These are contextual directives for a specific auditing step (Value-Add Audit) intended to focus the AI's analysis on code logic rather than documentation. They do not constitute an attempt to override system prompts or safety guardrails.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data (specifications and code provided by the user). However, the skill has no functional capabilities—it cannot execute commands, write to the filesystem, or make network requests. The lack of capability tiering makes it safe from exploitation via indirect injection.
- [EXTERNAL_DOWNLOADS]: The skill references other related skills using relative paths (e.g.,
../asymmetric-wins/SKILL.md), which is standard for local skill organization and does not involve remote code retrieval.
Audit Metadata