skills/qredence/skills/rlm-memory/Gen Agent Trust Hub

rlm-memory

Pass

Audited by Gen Agent Trust Hub on Jul 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the DaytonaInterpreter.execute() method to run Python code blocks within a sandbox. This is the intended mechanism for performing file I/O on the durable volume.
  • [PROMPT_INJECTION]: The skill facilitates long-term memory storage, which introduces a surface for indirect prompt injection if the stored data contains untrusted content.
  • Ingestion points: Data is ingested via load_text_from_volume(), load_from_volume(), get_buffer(), and reading files from the /home/daytona/memory/ path.
  • Boundary markers: The implementation examples do not use delimiters or instructions to ignore embedded content when processing retrieved memory content.
  • Capability inventory: The skill leverages interp.execute() which allows arbitrary Python code execution within the sandbox environment.
  • Sanitization: No evidence of input validation or sanitization is shown for data retrieved from persistent storage.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 12, 2026, 02:01 AM
Security Audit — agent-trust-hub — rlm-memory