desk-monitoring

Pass

Audited by Gen Agent Trust Hub on Aug 29, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and monitor real-time market data from external sources, including WebSockets and API endpoints. This represents a potential attack surface where malicious or anomalous market data could be crafted to influence the agent's interpretation, summaries, or alert conditions.
  • Ingestion points: Data is ingested via hyperliquid-websocket (allMids, l2Book, trades, etc.) and /info API calls.
  • Boundary markers: The skill instructs the agent to use "Facts with sources" and to label interpretation clearly on a single line, which provides some structural separation but does not fully isolate the model from malicious payloads in the data.
  • Capability inventory: The agent has the capability to write logs to /workspace/trading-desk/watch/ and generate alerts to the user and other bots.
  • Sanitization: There is no explicit mention of sanitizing or escaping the data fields before processing or inclusion in alerts.
  • [COMMAND_EXECUTION]: The instructions describe the creation and execution of polling scripts and supervisor processes within the /workspace/trading-desk/watch/ directory. While these are intended for monitoring purposes, the management and execution of local scripts by the agent represents a capability tier that should be monitored for unauthorized modifications.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 29, 2026, 09:32 AM
Security Audit — agent-trust-hub — desk-monitoring