shatter-and-recalibrate
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMPROMPT_INJECTIONNO_CODE
Full Analysis
- [Prompt Injection] (MEDIUM): The skill instructions use authoritative metaphors such as 'shatter compromised calibration' and 'rebuild from ground truth.' This language is characteristic of jailbreak prompts that attempt to invalidate the AI's existing safety alignment (viewed as 'compromised') in order to establish a new, potentially unrestricted operational mode ('ground truth').
- [No Code] (INFO): There are no scripts, binaries, or tool configurations included. The skill consists entirely of high-level instructions, making its actual behavior highly dependent on the agent's interpretation of the ambiguous 'primary protocol actions' and 'operational context'.
Audit Metadata