@tank/llm-app-patterns
Pass
Audited by Gen Agent Trust Hub on May 31, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of markdown documentation providing best practices for RAG systems, tool-using agents, and cost optimization. All code examples are provided as non-executable pseudocode for educational purposes.
- [NO_CODE]: There are no executable components, scripts, or binary files included in the skill package. All functionality is descriptive rather than active.
- [DATA_EXFILTRATION]: The skill manifest (
tank.json) explicitly disables outbound network access (network: { outbound: [] }), mitigating any risk of automated data transmission. - [COMMAND_EXECUTION]: Subprocess execution is disabled in the skill metadata (
subprocess: false). No shell commands or dynamic execution patterns were found in the instructions. - [CREDENTIALS_UNSAFE]: A thorough review of all documentation and configuration files found no hardcoded API keys, tokens, or other sensitive secrets.
Audit Metadata