best-practices-tau-dag
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists of markdown guidelines and agent configuration. It does not contain any executable scripts, binary files, or network-active code.
- [SAFE]: All examples and references, such as the repository
grahama1970/tau, are consistent with the skill's author context and purpose. No external or untrusted dependencies are introduced. - [SAFE]: The instructions emphasize security-conscious patterns for AI agents, including 'fail-closed' error handling, explicit model policies, and the use of immutable goal hashes to prevent unauthorized task modification.
- [SAFE]: No patterns of prompt injection, data exfiltration, or obfuscation were detected. The skill specifically discourages hiding orchestration policy in prose, which reduces the risk of ambiguity and manipulation.
Audit Metadata