best-minds-optimizer
Pass
Audited by Gen Agent Trust Hub on Mar 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates as a structural pre-processing layer for LLM prompts. It does not include external scripts, network operations, or file system access commands. All operations are confined to the natural language processing session.
- [PROMPT_INJECTION]: While the skill's primary function is to rewrite and execute modified prompts, it implements a mandatory 'Review Gate' (described in SKILL.md and references/optimize.md). This mechanism forces the agent to pause and display the rewritten prompt, expert persona, and reasoning framework for user approval before execution. This design prevents 'silent' or malicious reframing of user intent.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data to drive the optimization pipeline. The risk of the user input hijacking the agent's behavior is mitigated by the 'Review Gate' and specific instructions in the 'Follow-up Handling' section of SKILL.md, which limit the agent to the established expert framework and discourage unnecessary re-optimization.
- [COMMAND_EXECUTION]: The term 'execution' used throughout the documentation refers to the processing of a rewritten prompt by the LLM, not the execution of system-level commands. No subprocess calls or executable script patterns were detected.
Audit Metadata