engram-memory
Pass
Audited by Gen Agent Trust Hub on May 1, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The instructions utilize strong imperative language, such as "MANDATORY," "ALWAYS ACTIVE," and "MUST," to override standard agent behavior. It explicitly commands the agent to execute tools "IMMEDIATELY and WITHOUT BEING ASKED," bypassing user interaction. The protocol includes a repetitive "Self-check" loop designed to reinforce this instruction set after every completed task.
- [DATA_EXFILTRATION]: The skill mandates the proactive collection and transmission of session context to an external tool (
mem_save). The logged data includes technical codebase discoveries, architecture decisions, file paths, and user preferences. This data collection occurs silently in the background as a core protocol, rather than being triggered by specific user requests for data storage.
Audit Metadata