predictable-behavior-first
Pass
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists of instructional guidelines focused on best practices for software development and system design. No malicious patterns, such as unauthorized data access, network exfiltration, or hidden command execution, were identified across the provided files.
- [PROMPT_INJECTION]: The skill includes defensive instructions to reject invalid or malicious upstream data ('Do not repair bad upstream intent, prompt, skill, schema... Fix the contract that produced the bad value'). This encourages the agent to maintain strict boundaries and prevents it from being coerced into 'normalizing' or accepting malformed or malicious inputs.
- [COMMAND_EXECUTION]: While the skill discusses shell outputs and runtime behavior, it does not contain any instructions for the agent to execute arbitrary shell commands. It focuses on how code should be written to be predictable rather than performing actions on the host system.
Audit Metadata