cynic-judge
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWNO_CODE
Full Analysis
- No Code (SAFE): The skill consists of two Markdown files containing only instructions and reference data for the AI. There are no executable scripts, shell commands, or package dependencies.
- Indirect Prompt Injection (LOW): The skill's primary function is to evaluate external content (code, decisions, or project data). While this content could theoretically contain instructions intended to manipulate the evaluation, the skill lacks any side-effect capabilities (such as network access, file writing, or command execution), limiting the risk to the content of the AI's textual response. The rigid 36-dimension framework and 61.8% confidence cap serve as inherent constraints against simple manipulation.
- Data Exposure & Exfiltration (SAFE): No credentials, sensitive file paths, or network exfiltration patterns were detected. The skill operates entirely within the text processing context of the agent.
Audit Metadata