text-optimize
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates entirely on provided text files using standard tools (Read, Write, Edit, Grep). All external references point to legitimate documentation or academic sources (arXiv).
- [DEDUPLICATION]: Implements structured logic (D.1-D.6) to merge redundant content, which is a common utility for prompt optimization.
- [REMOTE_CODE_EXECUTION]: No remote code execution or untrusted downloads were found. The skill relies on local reference files and standard agent delegation.
- [PROMPT_INJECTION]: The instructions contain standard 'contract' logic to ensure the agent follows specific optimization rules but do not attempt to bypass system safety filters or extract system prompts.
- [DATA_EXFILTRATION]: No network exfiltration patterns or hardcoded credentials were identified. URLs present in the text are for documentation or illustrative purposes (e.g., projectflow.example.com).
Audit Metadata