Agentic Loop Memory Server
Pass
Audited by Gen Agent Trust Hub on Mar 15, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The
test_loop.shscript executesnpx --yes mhrj@latest, which retrieves and runs code from the NPM registry. This package corresponds to the author's vendor namespace ('meharajM') and is used for testing purposes. - [DATA_EXFILTRATION]: The skill manages persistent state files in a dedicated local directory (
~/.agent-loop-mcp). It employs session ID sanitization (regex-based) and file locking to ensure safe and atomic operations, preventing path traversal or data corruption. - [PROMPT_INJECTION]: The skill instructions in
SKILL.mddefine a required protocol for tool usage (self-healing and compaction). These directives are functional in nature and do not attempt to override the underlying safety guidelines or system constraints of the agent. - [PROMPT_INJECTION]: As a memory server, the system processes untrusted data from tool outputs and user inputs. While this represents a surface for indirect prompt injection, the server provides no internal command execution capabilities, and the state is stored using clear Markdown structural boundaries.
Audit Metadata