memory-manage

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill manages a persistent memory layer by ingesting project files and session history. This workflow creates an indirect prompt injection surface where external data is incorporated into the agent's long-term context.
  • Ingestion points: Files within the project and the ai-context/ directory.
  • Boundary markers: None detected; the instructions lack explicit delimiters to isolate data from instructions during the update process.
  • Capability inventory: The skill documentation describes tools for reading and writing local files to maintain context (mem_update, mem_pin, mem_review).
  • Sanitization: No sanitization or filtering logic is specified for the data being persisted into memory.
  • [SAFE]: The skill exclusively interacts with local project files and standard agent configuration paths. No unauthorized network activity or hardcoded credentials were identified.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 10:01 AM
Security Audit — agent-trust-hub — memory-manage