ai-safety-reviewer
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: No security issues were detected. The skill consists entirely of instructional markdown and configuration files aimed at improving AI safety and risk assessment.
- [PROMPT_INJECTION]: The skill mentions prompt injection and jailbreaks as failure modes to be analyzed in target systems, but does not contain instructions that override its own safety guidelines. It ingests system descriptions for review, which constitutes an indirect prompt injection surface.
- Ingestion points: Typical inputs defined in
SKILL.mdandmeta/skill.json, such as system descriptions, model capabilities, and tool access details. - Boundary markers: The instructions require defining a strict system context and harm surface before processing data, which creates a logical separation for analysis.
- Capability inventory: The skill has no capability to execute shell commands, write to the file system, or initiate network requests.
- Sanitization: There is no explicit input sanitization provided, as the skill's primary function is qualitative analysis.
- [DATA_EXPOSURE]: The skill does not access local credentials, environment variables, or sensitive user files.
Audit Metadata