memory-processor

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from agent_file_system/EVENT_UNPROCESSED.md, which contains raw transcripts of user messages.
  • [INDIRECT_PROMPT_INJECTION]: The processing workflow lacks specific boundary markers or sanitization instructions to prevent the agent from being influenced by malicious instructions embedded within user messages during the distillation process.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses file-writing capabilities through the stream_edit tool, allowing it to record potentially malicious content into agent_file_system/MEMORY.md, which serves as the agent's long-term memory store.
  • [INDIRECT_PROMPT_INJECTION]: Although the 'Distill, Don't Copy' principle and word count limits reduce the surface area for injection, an adversary can still craft messages designed to be summarized into persistent, harmful 'facts' or 'preferences' that affect future agent interactions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 01:51 PM
Security Audit — agent-trust-hub — memory-processor