iblai-api-agent-memory

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill facilitates the ingestion and retrieval of user-provided content into the agent's long-term memory (PGVector), creating a surface for indirect prompt injection.
  • Ingestion points: Memory content is ingested via POST requests to mentor-memories and global-memories endpoints defined in SKILL.md.
  • Boundary markers: According to references/concepts.md, memories are injected into the system prompt using Markdown headers (e.g., ## Relevant Context from Previous Conversations) to separate memory data from instructions.
  • Capability inventory: The skill provides full CRUD (Create, Read, Update, Delete) capabilities over the memory store via the platform API.
  • Sanitization: The system employs a MemoryExtractionService which uses an LLM (e.g., gpt-4o-mini) to filter and extract relevant memories, acting as a natural language validator before storage. However, malicious instructions could still persist if they appear relevant to the extraction prompt.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 11:38 AM
Security Audit — agent-trust-hub — iblai-api-agent-memory