The Agent Skills Directory

[SAFE]: The skill consists of behavioral instructions for optimizing model performance and token consumption. It does not contain any executable scripts, network requests, or attempts to access sensitive system information.
[INDIRECT_PROMPT_INJECTION]: The instructions describe how the agent should process external inputs like log files and source code (Rule 4). This creates a surface for indirect prompt injection if the files being processed contain malicious instructions. However, the skill's primary focus is on pruning and filtering this data, which reduces the potential for such attacks. The capability to read these files depends on the agent's pre-existing tools, not on code provided by this skill.

zipai-optimizer