Skip to content

fix(ENG-12470): update inline docs to match implementation, scope Tier 2 to Tier 1 risky fields by default #30

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
hiskudin merged 3 commits into from
Mar 27, 2026
Merged

fix(ENG-12470): update inline docs to match implementation, scope Tier 2 to Tier 1 risky fields by default #30

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
4b4ff7b
docs: remove redundant enableTier2: true from example (default is true)
hiskudin Mar 26, 2026
1e07e25
Merge branch 'main' into fix/tier2-default-example
hiskudin Mar 26, 2026
349ae6a
feat(ENG-12439): scope Tier 2 to Tier 1 risky fields by default
hiskudin Mar 26, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 9 additions & 3 deletions src/core/prompt-defense.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -127,7 +127,7 @@ export interface PromptDefenseOptions {
* ```typescript
* import { createPromptDefense } from '@stackone/defender';
*
* const defense = createPromptDefense({ enableTier2: true });
* const defense = createPromptDefense();
* await defense.warmupTier2();
*
* const result = await defense.defendToolResult(toolOutput, 'gmail_get_message');
Expand Down Expand Up @@ -226,7 +226,11 @@ export class PromptDefense {
let tier2Risk: RiskLevel = "low";

if (this.tier2Classifier) {
const strings = extractStrings(value, this.tier2Fields);
// Use explicit tier2Fields override, or fall back to the risky field names
// identified by Tier 1. If neither is available, scan all strings.
const { riskyFieldNames } = sanitized.metadata;
const fieldsForTier2 = this.tier2Fields ?? (riskyFieldNames.length > 0 ? riskyFieldNames : undefined);
const strings = extractStrings(value, fieldsForTier2);
const combinedText = strings.join("\n\n");

if (combinedText.length > 0) {
Expand All @@ -241,7 +245,9 @@ export class PromptDefense {
} else {
tier2SkipReason = this.tier2Fields?.length
? "No strings found in tier2Fields"
: "No strings extracted from tool result";
: riskyFieldNames.length > 0
? "No strings found in Tier 1 risky fields"
: "No strings extracted from tool result";
}
}

Expand Down
3 changes: 3 additions & 0 deletions src/core/tool-result-sanitizer.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -154,6 +154,7 @@ export class ToolResultSanitizer {
cumulativeRiskEscalated: false,
totalLatencyMs: 0,
sizeMetrics,
riskyFieldNames: [],
};

// Sanitize the value
Expand All @@ -167,6 +168,7 @@ export class ToolResultSanitizer {

metadata.totalLatencyMs = performance.now() - startTime;
metadata.sizeMetrics = sizeMetrics;
metadata.riskyFieldNames = [...new Set(metadata.riskyFieldNames)];

return {
sanitized: sanitized as T,
Expand Down Expand Up @@ -304,6 +306,7 @@ export class ToolResultSanitizer {

// Check if this is a risky field that needs sanitization
if (this.isFieldRisky(key, context.toolName) && typeof val === "string") {
metadata.riskyFieldNames.push(key);
result[key] = this.sanitizeStringField(val, fieldContext, toolRule, metadata);
} else {
// Recurse into non-risky fields
Expand Down
2 changes: 2 additions & 0 deletions src/types.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -217,6 +217,8 @@ export interface SanitizationMetadata {
totalLatencyMs: number;
/** Size metrics */
sizeMetrics: SizeMetrics;
/** Unique field names (leaf keys) that Tier 1 identified as risky */
riskyFieldNames: string[];
}

/**
Expand Down
Loading