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

feat(action): support optional Cyber access program selection (#1516)

dans-oai8ff5b39e1ecd
diff --git a/README.md b/README.md
index 1a1ea882f..441958068 100644
--- a/README.md
+++ b/README.md
@@ -21,10 +21,15 @@ fixing security vulnerabilities in your code.
Requires Node.js 22.13.0+ within 22.x, or Node.js 24.x or 26.x, and Python 3.10+.
Python 3.10 also requires `tomli`.
+This example explicitly requests Daybreak Blue and requires the corresponding
+access. Without Daybreak access, omit `--cyber-access-program daybreak_blue`
+or use `--cyber-access-program standard`.
+
```bash
npm install @openai/codex-security
npx @openai/codex-security login
-npx @openai/codex-security scan /path/to/repository
+npx @openai/codex-security scan /path/to/repository \
+ --cyber-access-program daybreak_blue
```
For CI, set `OPENAI_API_KEY` or `CODEX_API_KEY` in the scan process's environment.
@@ -104,6 +109,10 @@ repository scan, add an OpenAI API key as the repository secret
`.github/workflows/codex-security.yml`. Replace `REPLACE_WITH_REVIEWED_COMMIT`
with the full SHA of an Action commit.
+This workflow requests Daybreak Blue. Use an API key from a project with Blue
+enabled; without Daybreak access, omit `cyber-access-program` or set it to
+`standard`.
+
```yaml
name: Codex Security
on:
@@ -118,6 +127,7 @@ jobs:
security:
runs-on: ubuntu-24.04
steps:
+ # Configure Bubblewrap and AppArmor so Codex Security can run safely in its sandbox.
- name: Set up the Ubuntu sandbox
run: |
sudo apt-get update
@@ -130,6 +140,7 @@ jobs:
with:
model: gpt-5.6-sol
effort: high
+ cyber-access-program: daybreak_blue
env:
OPENAI_API_KEY: ${{ secrets.CODEX_SECURITY_API_KEY }}
```
diff --git a/action.yml b/action.yml
index 0ee14dcf0..bc42712e0 100644
--- a/action.yml
+++ b/action.yml
@@ -20,6 +20,8 @@ inputs:
model:
description: Model with access through your API key. Cost limits require CLI pricing support for the model.
default: gpt-5.6-sol
+ cyber-access-program:
+ description: 'Optional Cyber access program: standard, daybreak_blue, or daybreak_red. Unset preserves CLI defaults. Daybreak selections require approved organization and enabled API project access; model access is checked separately.'
effort:
description: 'Reasoning effort: minimal, low, medium, high, xhigh, or max (subject to model support).'
default: xhigh
diff --git a/github-action/README.md b/github-action/README.md
index 0099f2886..93117d0e3 100644
--- a/github-action/README.md
+++ b/github-action/README.md
@@ -15,6 +15,10 @@ save this workflow in `.github/workflows/codex-security.yml`.
Replace `REPLACE_WITH_REVIEWED_COMMIT` with the full SHA of an Action commit.
When using a fork, replace `openai` with the fork owner.
+This example requests Daybreak Blue. Use an API key from a project with Blue
+enabled; without Daybreak access, omit `cyber-access-program` or set it to
+`standard`. See [Daybreak access and refusals](#daybreak-access-and-refusals).
+
```yaml
name: Codex Security repository
on:
@@ -29,6 +33,7 @@ jobs:
security:
runs-on: ubuntu-24.04
steps:
+ # Configure Bubblewrap and AppArmor so Codex Security can run safely in its sandbox.
- name: Set up the Ubuntu sandbox
run: |
sudo apt-get update
@@ -41,6 +46,7 @@ jobs:
with:
model: gpt-5.6-sol
effort: high
+ cyber-access-program: daybreak_blue
env:
OPENAI_API_KEY: ${{ secrets.CODEX_SECURITY_API_KEY }}
```
@@ -51,6 +57,48 @@ Findings are report-only by default. Valid partial results produce a warning;
scanner and required reporting errors fail the job. Set `fail-on-severity` to
fail on findings at or above a selected severity when the scan is complete.
+### Ubuntu sandbox setup
+
+Codex Security runs commands inside a sandbox to limit their filesystem and
+network access. On the Ubuntu 24.04 runner used above, that sandbox uses
+Bubblewrap (`bwrap`), and Ubuntu's AppArmor policy must allow it to create the
+required isolated environment.
+
+The setup step installs `bubblewrap` and `apparmor-profiles`, then uses
+`apparmor_parser -r` to load the supplied `bwrap-userns-restrict` profile. The
+profile permits Bubblewrap's sandbox setup while retaining Ubuntu's broader
+AppArmor restrictions. Keep this step before the Action: the CLI checks that
+its sandbox works before scanning and fails if it cannot start, for example
+with `bwrap: loopback: Failed RTM_NEWADDR: Operation not permitted`.
+
+This setup is specific to the Linux runner. Local macOS scans use the built-in
+Seatbelt sandbox and do not need Bubblewrap. See the
+[sandbox documentation](https://learn.chatgpt.com/docs/sandboxing) for details.
+
+## Daybreak access and refusals
+
+Daybreak is optional. You can use the Action with a general-purpose model your
+API key can access under standard safeguards. Some security tasks may trigger
+model refusals or API cybersecurity safety blocks. **If a refusal or safety
+block aborts the scan, the Action fails**, even with `fail-on-severity: none`.
+That setting controls failures caused by findings; it does not suppress scanner
+errors. Check the job summary and logs for the reason. A failed scan does not
+mean the repository has no vulnerabilities.
+
+For authorized security work that needs reduced refusals,
+[request Daybreak access](https://openai.com/form/enterprise-trusted-access-for-cyber/).
+After organization approval, enable the required program for the API project
+and use a key from that project. Daybreak reduces some refusals but does not
+remove every safeguard or guarantee that a scan will complete.
+
+To explicitly select an approved Cyber access program, add
+`cyber-access-program: daybreak_blue` to the scan step's `with` settings.
+The input also accepts `standard` and `daybreak_red`; choose a program supported
+by your model and API project. Leaving it unset preserves CLI defaults.
+Selecting `standard` does not require Daybreak approval. Selecting a Daybreak
+program does not grant access; an unavailable program or incompatible model
+causes the scan to fail.
+
## Scan pull requests
Use the same API-key secret and Action commit as above. Check out the PR head
@@ -76,6 +124,7 @@ jobs:
runs-on: ubuntu-24.04
timeout-minutes: 60
steps:
+ # Configure Bubblewrap and AppArmor so Codex Security can run safely in its sandbox.
- name: Set up the Ubuntu sandbox
run: |
sudo apt-get update
@@ -277,25 +326,26 @@ validation. Verify report compatibility when adopting a new release.
Inputs are strings. Quote booleans and use newline-separated literal paths for lists.
-| Input | Default | Meaning |
-| ------------------ | ------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------- |
-| `repository` | `${{ github.workspace }}` | Checkout root. Use paths to select folders within the checkout. |
-| `scope` | `repository` | repository or diff. Select diff for PR changes only; repository scans the full checkout. |
-| `paths` | Unset | Newline-delimited literal repository-relative files or folders. Only for repository scope; no glob expansion. |
-| `diff-base` | Unset | Diff base revision. Defaults to the PR merge base; required outside PRs when scope is diff. |
-| `mode` | `standard` | standard or deep. Deep supports repository scans, including selected paths; not diff scans. |
-| `model` | `gpt-5.6-sol` | Model with access through your API key. Cost limits require CLI pricing support for the model. |
-| `effort` | `xhigh` | Reasoning effort: minimal, low, medium, high, xhigh, or max (subject to model support). |
-| `max-cost` | Unset | Positive estimated USD stop threshold per invocation. In-flight requests can exceed it; unset means no cost limit. |
-| `max-time-hours` | Unset | Positive Deep discovery duration in hours, up to 96. Unset uses the CLI default. Finalization and job timeout are separate. |
-| `fail-on-severity` | `none` | none, low, medium, high, or critical. Applies to complete scans. Valid partial results warn; scanner and required reporting errors fail. |
-| `verbose` | `true` | Stream CLI diagnostics to the job log. Set false for lifecycle and elapsed-time messages only. |
-| `dry-run` | `false` | Validate local configuration without a scan or API key. Does not verify authentication or model access. Use a separate non-required job. |
-| `summary` | `true` | Write a human-readable job summary. |
-| `annotations` | `true` | Emit up to 50 source finding annotations; complete findings remain in reports. |
-| `upload-artifacts` | `false` | Upload an allowlist of validated reports. Reports may contain source and vulnerability details. |
-| `artifact-name` | `codex-security` | Report artifact name; choose distinct names for matrix jobs and multiple invocations. |
-| `retention-days` | `7` | Artifact retention, 1–90 days (subject to repository limits). |
+| Input | Default | Meaning |
+| ---------------------- | ------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `repository` | `${{ github.workspace }}` | Checkout root. Use paths to select folders within the checkout. |
+| `scope` | `repository` | repository or diff. Select diff for PR changes only; repository scans the full checkout. |
+| `paths` | Unset | Newline-delimited literal repository-relative files or folders. Only for repository scope; no glob expansion. |
+| `diff-base` | Unset | Diff base revision. Defaults to the PR merge base; required outside PRs when scope is diff. |
+| `mode` | `standard` | standard or deep. Deep supports repository scans, including selected paths; not diff scans. |
+| `model` | `gpt-5.6-sol` | Model with access through your API key. Cost limits require CLI pricing support for the model. |
+| `cyber-access-program` | Unset | Optional Cyber access program: standard, daybreak_blue, or daybreak_red. Unset preserves CLI defaults. Daybreak selections require approved organization and enabled API project access; model access is checked separately. |
+| `effort` | `xhigh` | Reasoning effort: minimal, low, medium, high, xhigh, or max (subject to model support). |
+| `max-cost` | Unset | Positive estimated USD stop threshold per invocation. In-flight requests can exceed it; unset means no cost limit. |
+| `max-time-hours` | Unset | Positive Deep discovery duration in hours, up to 96. Unset uses the CLI default. Finalization and job timeout are separate. |
+| `fail-on-severity` | `none` | none, low, medium, high, or critical. Applies to complete scans. Valid partial results warn; scanner and required reporting errors fail. |
+| `verbose` | `true` | Stream CLI diagnostics to the job log. Set false for lifecycle and elapsed-time messages only. |
+| `dry-run` | `false` | Validate local configuration without a scan or API key. Does not verify authentication or model access. Use a separate non-required job. |
+| `summary` | `true` | Write a human-readable job summary. |
+| `annotations` | `true` | Emit up to 50 source finding annotations; complete findings remain in reports. |
+| `upload-artifacts` | `false` | Upload an allowlist of validated reports. Reports may contain source and vulnerability details. |
+| `artifact-name` | `codex-security` | Report artifact name; choose distinct names for matrix jobs and multiple invocations. |
+| `retention-days` | `7` | Artifact retention, 1–90 days (subject to repository limits). |
## Outputs
diff --git a/github-action/dist/index.cjs b/github-action/dist/index.cjs
index a5a8cbd97..be85b9b37 100644
--- a/github-action/dist/index.cjs
+++ b/github-action/dist/index.cjs
@@ -55585,12 +55585,14 @@ function parseInputs(read, workspace) {
const maxTimeHours = num("max-time-hours", false, Number.MIN_VALUE, 96);
if (maxTimeHours !== void 0 && mode !== "deep") throw new Error("max-time-hours requires mode: deep.");
const dryRun = bool("dry-run", false);
+ const cyberAccessProgram = str("cyber-access-program") ? choice("cyber-access-program", ["standard", "daybreak_blue", "daybreak_red"], "standard") : void 0;
return {
repository: str("repository", workspace),
scope,
paths,
diffBase,
mode,
+ cyberAccessProgram,
model: str("model", "gpt-5.6-sol"),
effort: choice("effort", ["minimal", "low", "medium", "high", "xhigh", "max"], "xhigh"),
maxCost: num("max-cost", false, Number.MIN_VALUE),
@@ -55627,6 +55629,7 @@ function scanArguments(inputs, target, resultsDirectory) {
args.push("--codex", "analytics.enabled=false");
for (const path6 of inputs.paths) args.push(`--path=./${path6}`);
const options = [
+ ["--cyber-access-program", inputs.cyberAccessProgram],
["--diff", target.diffBase],
["--head", target.diffHead],
["--max-cost", inputs.maxCost],
@@ -99279,6 +99282,7 @@ async function runAction(actionRoot, overrides = {}) {
log2(`Target commit: ${target.scannedSha}.${target.diffBase ? ` Diff: ${target.diffBase}..${target.diffHead}.` : ""}`);
if (inputs.paths.length) log2(`Paths: ${inputs.paths.join(", ")}.`);
log2(`Model: ${inputs.model}; estimated cost stop threshold: ${inputs.maxCost === void 0 ? "unset" : `$${inputs.maxCost}`}; findings failure threshold: ${inputs.failOnSeverity}.`);
+ if (inputs.cyberAccessProgram) log2(`Requested Cyber access program: ${inputs.cyberAccessProgram}.`);
const scanLabel = inputs.dryRun ? "CLI configuration validation" : "Security scan";
const scanStarted = performance.now();
info(`Starting ${scanLabel.toLowerCase()}. CLI diagnostics: ${inputs.verbose ? "streaming" : "disabled (verbose: false)"}.`);
diff --git a/github-action/scripts/cli-integration.ts b/github-action/scripts/cli-integration.ts
index 869760284..55f0b34c6 100644
--- a/github-action/scripts/cli-integration.ts
+++ b/github-action/scripts/cli-integration.ts
@@ -81,6 +81,16 @@ try {
assert.equal(deepPreflight.maxTimeHours, 0.25);
assert.deepEqual(deepPreflight.target.paths, ['example.ts']);
+ // Validate explicit program selection against the locked CLI, without model calls.
+ for (const mode of ['standard', 'deep']) for (const program of ['standard', 'daybreak_blue', 'daybreak_red']) {
+ const values: Record<string, string> = {mode, 'cyber-access-program':program, 'dry-run':'true'};
+ const args = scanArguments(parseInputs(name => values[name] ?? '', repository),
+ {repository}, join(root, `cyber-${mode}-${program}`));
+ const preflight = JSON.parse(run(args, 0));
+ assert.equal(preflight.dryRun, true);
+ assert.equal(preflight.mode, mode);
+ }
+
// Normalization must not let a literal repository path become a framework option.
await mkdir(join(repository, '--help'));
await writeFile(join(repository, '--help/example.ts'), 'export const example = 1;\n');
diff --git a/github-action/src/inputs.ts b/github-action/src/inputs.ts
index 8c0dc6e48..97c3320c0 100644
--- a/github-action/src/inputs.ts
+++ b/github-action/src/inputs.ts
@@ -1,6 +1,6 @@
export const INPUT_NAMES = [
'repository', 'scope', 'paths', 'diff-base', 'mode',
- 'model', 'effort', 'max-cost', 'max-time-hours', 'fail-on-severity', 'verbose', 'dry-run',
+ 'model', 'cyber-access-program', 'effort', 'max-cost', 'max-time-hours', 'fail-on-severity', 'verbose', 'dry-run',
'summary', 'annotations',
'upload-artifacts', 'artifact-name', 'retention-days',
] as const;
@@ -42,8 +42,11 @@ export function parseInputs(read: (name: string) => string, workspace: string) {
const maxTimeHours = num('max-time-hours', false, Number.MIN_VALUE, 96);
if (maxTimeHours !== undefined && mode !== 'deep') throw new Error('max-time-hours requires mode: deep.');
const dryRun = bool('dry-run', false);
+ const cyberAccessProgram = str('cyber-access-program')
+ ? choice('cyber-access-program', ['standard', 'daybreak_blue', 'daybreak_red'], 'standard')
+ : undefined;
return {
- repository: str('repository', workspace), scope, paths, diffBase, mode,
+ repository: str('repository', workspace), scope, paths, diffBase, mode, cyberAccessProgram,
model: str('model', 'gpt-5.6-sol'), effort: choice('effort', ['minimal','low','medium','high','xhigh','max'], 'xhigh'),
maxCost: num('max-cost', false, Number.MIN_VALUE), maxTimeHours, failOnSeverity: choice('fail-on-severity', ['none','low','medium','high','critical'], 'none'),
verbose: bool('verbose', true), dryRun,
@@ -66,6 +69,7 @@ export function scanArguments(inputs: Inputs, target: {repository: string; diffB
// leading '-' and '~' as literal repository filenames.
for (const path of inputs.paths) args.push(`--path=./${path}`);
const options: Array<[string, string | number | undefined]> = [
+ ['--cyber-access-program', inputs.cyberAccessProgram],
['--diff', target.diffBase], ['--head', target.diffHead], ['--max-cost', inputs.maxCost],
['--max-time-hours', inputs.maxTimeHours],
];
diff --git a/github-action/src/main.ts b/github-action/src/main.ts
index 1020e7477..fe62d854e 100644
--- a/github-action/src/main.ts
+++ b/github-action/src/main.ts
@@ -97,6 +97,7 @@ export async function runAction(actionRoot: string, overrides: Partial<Dependenc
log(`Target commit: ${target.scannedSha}.${target.diffBase ? ` Diff: ${target.diffBase}..${target.diffHead}.` : ''}`);
if (inputs.paths.length) log(`Paths: ${inputs.paths.join(', ')}.`);
log(`Model: ${inputs.model}; estimated cost stop threshold: ${inputs.maxCost === undefined ? 'unset' : `$${inputs.maxCost}`}; findings failure threshold: ${inputs.failOnSeverity}.`);
+ if (inputs.cyberAccessProgram) log(`Requested Cyber access program: ${inputs.cyberAccessProgram}.`);
const scanLabel = inputs.dryRun ? 'CLI configuration validation' : 'Security scan';
const scanStarted = performance.now();
core.info(`Starting ${scanLabel.toLowerCase()}. CLI diagnostics: ${inputs.verbose ? 'streaming' : 'disabled (verbose: false)'}.`);
diff --git a/github-action/tests/inputs.test.ts b/github-action/tests/inputs.test.ts
index 500a4bbcf..b862b7634 100644
--- a/github-action/tests/inputs.test.ts
+++ b/github-action/tests/inputs.test.ts
@@ -6,7 +6,7 @@ const parse = (values: Record<string, string> = {}) => parseInputs(key => values
test('defaults are stable across events', () => {
const input = parse();
assert.deepEqual(input, {
- repository: '/checkout', scope: 'repository', paths: [], diffBase: undefined, mode: 'standard',
+ repository: '/checkout', scope: 'repository', paths: [], diffBase: undefined, mode: 'standard', cyberAccessProgram: undefined,
model: 'gpt-5.6-sol', effort: 'xhigh', maxCost: undefined, maxTimeHours: undefined, failOnSeverity: 'none',
verbose: true, dryRun: false, summary: true, annotations: true,
uploadArtifacts: false, artifactName: 'codex-security', retentionDays: 7,
@@ -16,6 +16,16 @@ test('defaults are stable across events', () => {
assert.ok(!args.includes('--max-time-hours'));
assert.ok(!args.includes('--fail-on-severity'));
assert.ok(!args.includes('--python'));
+ assert.ok(!args.includes('--cyber-access-program'));
+});
+test('Cyber access selections reach both standard and Deep scan CLI invocations', () => {
+ for (const mode of ['standard', 'deep']) for (const program of ['standard', 'daybreak_blue', 'daybreak_red']) {
+ const args = scanArguments(parse({mode, 'cyber-access-program':program}), {repository:'/checkout'}, '/results');
+ assert.equal(args[args.indexOf('--cyber-access-program') + 1], program);
+ }
+ const blank = scanArguments(parse({'cyber-access-program':' '}), {repository:'/checkout'}, '/results');
+ assert.ok(!blank.includes('--cyber-access-program'));
+ assert.throws(() => parse({'cyber-access-program':'unsupported'}), /cyber-access-program must be one of/);
});
test('Deep scans forward an explicit discovery budget and selected repository paths', () => {
const input = parse({mode:'deep', paths:'./src/\nlib', 'max-time-hours':'1.5'});