Skip to content

OpenAI's Codex Security CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities. npm: https://www.npmjs.com/package/@openai/codex-security

refactor(sdk): share the severity review count default (#2479)

Michael D'Angelo4c4fff3897ad
diff --git a/sdk/typescript/src/classify-severity-records.ts b/sdk/typescript/src/classify-severity-records.ts
index 609f0c3c7..a518f46cd 100644
--- a/sdk/typescript/src/classify-severity-records.ts
+++ b/sdk/typescript/src/classify-severity-records.ts
@@ -6,7 +6,7 @@ import { abortable } from "./targets.js";
import {
parseSeverityDecision,
reviewSeverity,
- classificationRunsSchema,
+ reviewRunsOption,
severityDecisionSchema,
severityReviewInstructions,
type SeverityDecision,
@@ -128,9 +128,7 @@ export async function classifySeverityRecords(
options: ClassifySeverityRecordsOptions,
): Promise<ClassifySeverityRecordsResult> {
options.signal?.throwIfAborted();
- const classificationRuns = classificationRunsSchema
- .default(3)
- .parse(options.classificationRuns);
+ const classificationRuns = reviewRunsOption.parse(options.classificationRuns);
const data = classifySeverityRecordsInputSchema.parse(structuredClone(input));
const rubric = data.rubric ?? {
revision: 1,
diff --git a/sdk/typescript/src/classify-severity.ts b/sdk/typescript/src/classify-severity.ts
index 8ff9c77e2..08cfc903b 100644
--- a/sdk/typescript/src/classify-severity.ts
+++ b/sdk/typescript/src/classify-severity.ts
@@ -19,6 +19,7 @@ import {
parseSeverityDecision,
reviewSeverity,
classificationRunsSchema,
+ reviewRunsOption,
} from "./severity-review.js";
import { CODEX_SECURITY_THREAD_SOURCES } from "./thread-source.js";
@@ -124,9 +125,7 @@ export async function classifySeverityInternal(
checkpoint?: SeverityClassificationCheckpoint,
): Promise<SeverityClassification> {
options.signal?.throwIfAborted();
- const classificationRuns = classificationRunsSchema
- .default(3)
- .parse(options.classificationRuns);
+ const classificationRuns = reviewRunsOption.parse(options.classificationRuns);
const ids = new Set<string>();
for (const finding of findings) {
if (!finding.findingId?.trim() || ids.has(finding.findingId)) {
diff --git a/sdk/typescript/src/cli.ts b/sdk/typescript/src/cli.ts
index 7d7f6e537..009933f3f 100644
--- a/sdk/typescript/src/cli.ts
+++ b/sdk/typescript/src/cli.ts
@@ -1,5 +1,5 @@
#!/usr/bin/env node
-import { classificationRunsSchema } from "./severity-review.js";
+import { DEFAULT_REVIEW_RUNS, reviewRunsOption } from "./severity-review.js";
import { severityRubricSchema } from "./severity-rubric.js";
import { explicitlyCanceled } from "./scan-executor.js";
import { workbenchArgs } from "./workbench-arguments.js";
@@ -2275,12 +2275,11 @@ export async function main(
runs: { type: "string" },
},
});
- classificationRuns = classificationRunsSchema.parse(
- values.runs === undefined ? 3 : Number(values.runs),
- );
+ if (values.runs !== undefined)
+ classificationRuns = reviewRunsOption.parse(Number(values.runs));
} catch {
errorOutput.write(
- "codex-security: Use --records with only --runs (an integer from 1 to 10; default: 3).\n",
+ `codex-security: Use --records with only --runs (an integer from 1 to 10; default: ${DEFAULT_REVIEW_RUNS}).\n`,
);
return 2;
}
@@ -5438,11 +5437,9 @@ export async function main(
hint: "For saved findings, use --rubric for your own policy or set CODEX_SECURITY_ENABLE_REMOTE_SEVERITY_RUBRIC=1 for the Statsig default. For host-provided reviews, use --records.",
destructive: true,
options: z.object({
- runs: classificationRunsSchema
- .default(3)
- .describe(
- "Number of independent classification runs (1–10; default: 3).",
- ),
+ runs: reviewRunsOption.describe(
+ `Number of independent classification runs (1–10; default: ${DEFAULT_REVIEW_RUNS}).`,
+ ),
records: z
.boolean()
.default(false)
diff --git a/sdk/typescript/src/severity-review.ts b/sdk/typescript/src/severity-review.ts
index 3d8aa1d9c..fde4ea1d7 100644
--- a/sdk/typescript/src/severity-review.ts
+++ b/sdk/typescript/src/severity-review.ts
@@ -2,7 +2,10 @@ import { SCAN_SEVERITIES } from "./scan-settings.js";
import { z } from "zod";
import { CodexSecurityError } from "./errors.js";
+export const DEFAULT_REVIEW_RUNS = 3;
export const classificationRunsSchema = z.number().int().min(1).max(10);
+export const reviewRunsOption =
+ classificationRunsSchema.default(DEFAULT_REVIEW_RUNS);
const levelSchema = z.enum(SCAN_SEVERITIES);
const textSchema = z
@@ -64,9 +67,8 @@ export function parseSeverityDecision(
/** Run independent reviews and retain the first complete answer for the winning level. */
export async function reviewSeverity(
review: () => Promise<SeverityDecision>,
- classificationRuns = 3,
+ runs: number,
): Promise<SeverityDecision> {
- const runs = classificationRunsSchema.parse(classificationRuns);
const decisions: SeverityDecision[] = [];
for (let run = 0; run < runs; run++) decisions.push(await review());
// Exclusion is a distinct vote; any assessed severity wins a tied vote.
diff --git a/sdk/typescript/tests-ts/severity-review.test.ts b/sdk/typescript/tests-ts/severity-review.test.ts
index c7f3496f6..976ef2823 100644
--- a/sdk/typescript/tests-ts/severity-review.test.ts
+++ b/sdk/typescript/tests-ts/severity-review.test.ts
@@ -29,7 +29,7 @@ test("three independent reviews vote by normalized severity, with deterministic
reviewTrigger: `Missing fact ${index + 1}`,
}));
let calls = 0;
- const result = await reviewSeverity(async () => answers[calls++]!);
+ const result = await reviewSeverity(async () => answers[calls++]!, 3);
const expected =
first === second || first === third
? first