Skip to content

Graphorin API reference v0.7.0


Graphorin API reference / @graphorin/security / / deriveTaintLabel

Function: deriveTaintLabel()

ts
function deriveTaintLabel(input): TaintLabel;

Defined in: packages/security/src/dataflow/derive.ts:45

Derive the provenance label for a tool's output from its resolved trust class, source, and declared sensitivity.

  • untrusted is keyed off the ToolTrustClass: mcp-derived, web-search, and skill-untrusted produce untrusted output.
  • sensitive is true only for the 'secret' tier. 'internal' is the default tier for ordinary user content, so counting it would make the lethal-trifecta gate fire on essentially every run; operators who want a broader gate widen it via policy, not here.

Parameters

ParameterTypeDescription
input{ sensitiveTiers?: readonly Sensitivity[]; sensitivity?: Sensitivity; source?: ToolSource; trustClass: ToolTrustClass; }-
input.sensitiveTiers?readonly Sensitivity[]Sensitivity tiers that count as "sensitive" for the lethal-trifecta leg (SDF-8). Default ['secret'] - out of the box only secret-tagged content arms the trifecta, so the gate does not fire on every run. Widen to e.g. ['secret', 'internal'] to also treat ordinary user/PII content (which defaults to 'internal') as sensitive.
input.sensitivity?Sensitivity-
input.source?ToolSource-
input.trustClassToolTrustClass-

Returns

TaintLabel

Stable