hermes-hudui-consciousness-monitor

Warn

Audited by Gen Agent Trust Hub on May 17, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to clone a full application codebase from an external GitHub repository (https://github.com/joeynyc/hermes-hudui.git). This source is not part of the platform's trusted vendor list.
  • [COMMAND_EXECUTION]: The installation workflow requires the execution of a shell script (./install.sh) and a binary (hermes-hudui) downloaded from the remote repository. Executing unverified scripts from external sources poses a significant risk of arbitrary code execution.
  • [DATA_EXFILTRATION]: The skill configuration and code examples demonstrate broad access to sensitive files in the ~/.hermes/ directory, including identity.json (agent credentials/config), memory/ (long-term data), and sessions/ (interaction history). While used for local display, this access allows the tool to ingest all of the agent's private data.
  • [INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection by processing untrusted data.
  • Ingestion points: Reads agent session logs and memory stores from ~/.hermes/sessions/ and ~/.hermes/memory/ for display in the Web UI.
  • Boundary markers: The skill mentions a "Safe Share Mode" for redacted exports, but no specific boundaries are described for the live dashboard rendering to prevent embedded instructions from affecting the agent via the UI.
  • Capability inventory: The dashboard provides 'Gateway control' and 'Plugin management' (enable/disable), which could be manipulated if malicious data is rendered without sanitization.
  • Sanitization: The skill description does not specify sanitization or escaping methods for external content rendered within the dashboard interface.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 17, 2026, 12:32 AM
Security Audit — agent-trust-hub — hermes-hudui-consciousness-monitor