full-stop
Warn
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: MEDIUMPROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
- [PROMPT_INJECTION]: The skill contains instructions that explicitly direct the agent to override or ignore safety constraints defined in the target project's own files. It specifies that a 'full stop' request serves as authorization to bypass rules requiring the agent to 'ask first' before using certain test methods, instructing the agent to 'Use those methods directly.'
- [PRIVILEGE_ESCALATION]: By instructing the agent to ignore 'ask first' protocols found in project configuration or documentation, the skill attempts to elevate the agent's operational autonomy. This bypasses human-in-the-loop (HITL) safety checkpoints for potentially impactful actions such as computer use and automated testing.
Audit Metadata