@tank/agent-creator
Pass
Audited by Gen Agent Trust Hub on May 31, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill manifest (tank.json) implements a highly restricted environment, explicitly disabling network outbound traffic and subprocess execution.
- [SAFE]: The documentation and reference files (SKILL.md, references/*.md) provide architectural guidance for building agent systems. They emphasize security principles, specifically identity-first role definition to control behavior and least-privilege tool scoping to minimize risk.
- [SAFE]: The skill introduces a
readonlyfield in its agent schema, recommending it for analysis-focused agents to prevent accidental or malicious file mutations. - [SAFE]: No malicious patterns, such as prompt injections, hardcoded credentials, or obfuscated code, were found across any of the analyzed files.
Audit Metadata