Risk Scoring
How Faramesh models risk context alongside policy rules.
The checked-in source shows two complementary risk surfaces. In the Go CLI, faramesh policy simulate accepts --risk-score and injects risk_score into the simulated action args. In the SDK policy helpers, RiskLevel is an explicit enum with low, medium, and high values.
That means risk is part of the decision context, not a vague afterthought. The policy layer can allow an action while the surrounding runtime still escalates it into approval flow based on risk evidence.
from faramesh import RiskLevel, PolicyRule, MatchConditionexport type RiskLevel = "low" | "medium" | "high";In practice, risk in this codebase is driven by things the runtime already knows: tool class, operation, parameter shape, amount thresholds, agent identity, and sequence context.
See Risk Levels and Policy Simulation for the commands that expose the risk path.