agent-monitoring
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill interacts with the Monte Carlo platform using official vendor tools and domains. All external references point to the author's own infrastructure (getmontecarlo.com) or standard package registries, representing legitimate vendor functionality.- [EXTERNAL_DOWNLOADS]: The README provides setup instructions for installing the toolkit using npx from the author's GitHub repository (monte-carlo-data/mc-agent-toolkit), which is a standard procedure for this skill type.- [DATA_EXPOSURE]: The instructions explicitly guide the agent to protect sensitive internal identifiers by not exposing MCONs (Monte Carlo Object Names) or warehouse UUIDs to the end user during the agent discovery process.- [INDIRECT_PROMPT_INJECTION]: The skill ingests agent conversation traces and logs from the Monte Carlo platform. While these are external inputs that could potentially contain malicious instructions, the risk is minimized by the skill's specific focus on observability and the requirement for user confirmation.
- Ingestion points: The skill uses get_agent_conversation and get_agent_trace in SKILL.md to retrieve external trace data.
- Boundary markers: No specific delimiters or 'ignore' instructions are provided for the content of the conversation logs.
- Capability inventory: The agent can create monitoring configurations using specialized tools like create_agent_metric_monitor and create_agent_evaluation_monitor.
- Sanitization: The skill enforces a mandatory dry_run=True preview and explicit user confirmation before any monitor is actually created, providing a human-in-the-loop safety mechanism.
Audit Metadata