monte-carlo-context-detection
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's core functionality relies on processing untrusted data sources to determine routing logic. These ingestion points include user messages, workspace file names (e.g.,
dbt_project.yml,montecarlo.yml), and IDE context (e.g., active.sqlfiles). - [INDIRECT_PROMPT_INJECTION]: The skill performs 'auto-activation' of other workflows based on the signals gathered from these untrusted sources. The lack of explicit boundary markers or sanitization for file paths and table names before they are used in routing logic or passed to MCP tools creates a surface for potential schema confusion or indirect injection attacks.
- [INDIRECT_PROMPT_INJECTION]: The capability inventory includes the ability to load and execute other skill files (
SKILL.md) and call powerful data observability tools (get_alerts,search,get_monitors). If an attacker can craft environment signals (like specific file names) that influence the agent's confidence level, they may trigger unintended workflows or tool calls.
Audit Metadata