Skip to content

Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS

Rename the app PR tab's "PR review" action to "Untrusted:Review" (#12846)

[._.]/bff6ecddad89
diff --git a/client/src/components/apps/tabs/PullRequestsTab.jsx b/client/src/components/apps/tabs/PullRequestsTab.jsx
index 3eb41b4c9c..c84d8e5793 100644
--- a/client/src/components/apps/tabs/PullRequestsTab.jsx
+++ b/client/src/components/apps/tabs/PullRequestsTab.jsx
@@ -87,7 +87,7 @@ const ACTION_KINDS = {
&& Number(task.metadata?.reviewLoopPRNumber) === number,
},
review: {
- label: 'PR review',
+ label: 'Untrusted:Review',
verb: 'run the pr-reviewer task against',
Icon: ScanSearch,
field: 'reviewAction',
@@ -345,7 +345,7 @@ export default function PullRequestsTab({ appId, appName }) {
const [mergeMethod, setMergeMethod] = useState(DEFAULT_MERGE_METHOD);
const [deleteBranchOnMerge, setDeleteBranchOnMerge] = useState(true);
- // Resolve & merge and Do:Review default to the active provider. PR review
+ // Resolve & merge and Do:Review default to the active provider. Untrusted:Review
// follows its own scheduled stages unless the user explicitly opts this pin in.
// Auto remains available for server-side routing. Mirrors the Issues
// tab's "Run with" picker: a session convenience, never persisted.
@@ -662,7 +662,7 @@ export default function PullRequestsTab({ appId, appName }) {
</p>
{offersKind('review') && (
<p>
- PR review points the <span className="font-mono">pr-reviewer</span> scheduled task at this one request instead of letting it sweep every open contributor PR. It appears only on requests it can review — opened by someone else against the default branch — and uses the providers saved on its review stages, so the Code review setting below does not apply to it. The security scan must pass before either review stage runs.
+ Untrusted:Review points the <span className="font-mono">pr-reviewer</span> scheduled task at this one request instead of letting it sweep every open contributor PR. It appears only on requests it can review — opened by someone else against the default branch — and uses the providers saved on its review stages, so the Code review setting below does not apply to it. The security scan must pass before either review stage runs.
</p>
)}
{offersKind('doReview') && (
@@ -728,9 +728,9 @@ export default function PullRequestsTab({ appId, appName }) {
<div className="text-xs text-gray-400">
<input id="apply-run-with-to-pr-review" type="checkbox" checked={applyRunWithToPrReview}
onChange={event => setApplyRunWithToPrReview(event.target.checked)} className="mr-2" />
- <label htmlFor="apply-run-with-to-pr-review">Also run PR review on the Run with provider</label>
+ <label htmlFor="apply-run-with-to-pr-review">Also run Untrusted:Review on the Run with provider</label>
<p className="mt-1">
- Off by default: PR review is a <span className="font-mono">pr-reviewer</span> run, and that task keeps the providers saved on its own stages. Tick this to point its scan and review stages at the provider chosen above instead. Resolve &amp; merge and Do:Review always use it.
+ Off by default: Untrusted:Review is a <span className="font-mono">pr-reviewer</span> run, and that task keeps the providers saved on its own stages. Tick this to point its scan and review stages at the provider chosen above instead. Resolve &amp; merge and Do:Review always use it.
</p>
</div>
)}
diff --git a/client/src/components/apps/tabs/PullRequestsTab.test.jsx b/client/src/components/apps/tabs/PullRequestsTab.test.jsx
index 9d42864414..ce079e26a4 100644
--- a/client/src/components/apps/tabs/PullRequestsTab.test.jsx
+++ b/client/src/components/apps/tabs/PullRequestsTab.test.jsx
@@ -175,17 +175,17 @@ describe('PullRequestsTab', () => {
fireEvent.change(await screen.findByLabelText('Provider'), { target: { value: 'claude' } });
fireEvent.change(screen.getByLabelText('Model'), { target: { value: 'claude-opus-5' } });
- fireEvent.click(screen.getByRole('button', { name: /PR review/ }));
+ fireEvent.click(screen.getByRole('button', { name: /Untrusted:Review/ }));
await waitFor(() => expect(api.reviewAppPullRequest).toHaveBeenCalledWith(
'app-1', 17, {},
));
});
- // PR review hands its run to the `pr-reviewer` scheduled task, so it takes the
+ // Untrusted:Review hands its run to the `pr-reviewer` scheduled task, so it takes the
// provider pin only when the user opts in — and never the Code review setting,
// which governs the two actions PortOS composes here.
- it('passes the Run with pin to PR review once the opt-in is ticked, and nothing else', async () => {
+ it('passes the Run with pin to Untrusted:Review once the opt-in is ticked, and nothing else', async () => {
api.getProviders.mockResolvedValue({
providers: [{
id: 'claude', name: 'Claude', type: 'cli', enabled: true,
@@ -199,8 +199,8 @@ describe('PullRequestsTab', () => {
fireEvent.change(screen.getByLabelText('Model'), { target: { value: 'claude-opus-5' } });
fireEvent.change(screen.getByLabelText('Code review'), { target: { value: 'self' } });
- fireEvent.click(screen.getByLabelText('Also run PR review on the Run with provider'));
- fireEvent.click(screen.getByRole('button', { name: /PR review/ }));
+ fireEvent.click(screen.getByLabelText('Also run Untrusted:Review on the Run with provider'));
+ fireEvent.click(screen.getByRole('button', { name: /Untrusted:Review/ }));
await waitFor(() => expect(api.reviewAppPullRequest).toHaveBeenCalledWith(
'app-1', 17, { provider: 'claude', model: 'claude-opus-5', effort: undefined },
@@ -328,12 +328,12 @@ describe('PullRequestsTab', () => {
it('queues a pr-reviewer run scoped to the row it was pressed on', async () => {
await renderTab();
- fireEvent.click(await screen.findByRole('button', { name: /PR review/ }));
+ fireEvent.click(await screen.findByRole('button', { name: /Untrusted:Review/ }));
await waitFor(() => expect(api.reviewAppPullRequest).toHaveBeenCalledWith(
'app-1', 17, {},
));
- expect(await screen.findByRole('link', { name: /PR review: Queued/ })).toBeInTheDocument();
+ expect(await screen.findByRole('link', { name: /Untrusted:Review: Queued/ })).toBeInTheDocument();
// The resolve action is a separate lane and must stay offered.
expect(screen.getByRole('button', { name: /Resolve & merge/ })).toBeInTheDocument();
});
@@ -341,8 +341,8 @@ describe('PullRequestsTab', () => {
it('binds a pr-reviewer task update to the row by its target PR number', async () => {
await renderTab();
- fireEvent.click(await screen.findByRole('button', { name: /PR review/ }));
- expect(await screen.findByRole('link', { name: /PR review: Queued/ })).toBeInTheDocument();
+ fireEvent.click(await screen.findByRole('button', { name: /Untrusted:Review/ }));
+ expect(await screen.findByRole('link', { name: /Untrusted:Review: Queued/ })).toBeInTheDocument();
act(() => socketHandlers.get('cos:tasks:changed')({
task: {
@@ -352,7 +352,7 @@ describe('PullRequestsTab', () => {
},
}));
- expect(await screen.findByRole('link', { name: /PR review: Active/ })).toBeInTheDocument();
+ expect(await screen.findByRole('link', { name: /Untrusted:Review: Active/ })).toBeInTheDocument();
});
it('restores the preflight explanation after reloading the PR list', async () => {
@@ -361,13 +361,13 @@ describe('PullRequestsTab', () => {
}]));
await renderTab();
expect(await screen.findByRole('alert')).toHaveTextContent('Prompt Guard stopped');
- expect(screen.getByRole('button', { name: 'Retry PR review' })).toBeInTheDocument();
+ expect(screen.getByRole('button', { name: 'Retry Untrusted:Review' })).toBeInTheDocument();
});
it('replaces a queued review with a durable preflight explanation and retry', async () => {
await renderTab();
- fireEvent.click(await screen.findByRole('button', { name: /PR review/ }));
- await screen.findByRole('link', { name: /PR review: Queued/ });
+ fireEvent.click(await screen.findByRole('button', { name: /Untrusted:Review/ }));
+ await screen.findByRole('link', { name: /Untrusted:Review: Queued/ });
act(() => socketHandlers.get('cos:tasks:changed')({ task: {
id: 'preflight-17', status: 'completed',
metadata: { app: 'app-1', analysisType: 'pr-reviewer', targetPullRequest: 17,
@@ -375,9 +375,9 @@ describe('PullRequestsTab', () => {
} }));
expect(screen.getByRole('alert')).toHaveTextContent('before an agent started');
expect(screen.getByRole('link', { name: 'View failure record' })).toHaveAttribute('href', '/cos/tasks?task=preflight-17&source=internal');
- fireEvent.click(screen.getByRole('button', { name: 'Retry PR review' }));
+ fireEvent.click(screen.getByRole('button', { name: 'Retry Untrusted:Review' }));
await waitFor(() => expect(api.reviewAppPullRequest).toHaveBeenCalledTimes(2));
- expect(await screen.findByRole('link', { name: /PR review: Queued/ })).toBeInTheDocument();
+ expect(await screen.findByRole('link', { name: /Untrusted:Review: Queued/ })).toBeInTheDocument();
});
// #7258: the preflight card is the programmatic phase of this row's run, and
@@ -386,14 +386,14 @@ describe('PullRequestsTab', () => {
// later update, because a row binds to one task id for good.
it('follows a finished preflight card onto the review task it started', async () => {
await renderTab();
- fireEvent.click(await screen.findByRole('button', { name: /PR review/ }));
- await screen.findByRole('link', { name: /PR review: Queued/ });
+ fireEvent.click(await screen.findByRole('button', { name: /Untrusted:Review/ }));
+ await screen.findByRole('link', { name: /Untrusted:Review: Queued/ });
act(() => socketHandlers.get('cos:tasks:changed')({ task: {
id: 'preflight-demand-1', status: 'in_progress',
metadata: { app: 'app-1', analysisType: 'pr-reviewer', targetPullRequest: 17,
preflight: { phase: 'preparing', steps: [] } },
} }));
- expect(await screen.findByRole('link', { name: /PR review: Active/ })).toBeInTheDocument();
+ expect(await screen.findByRole('link', { name: /Untrusted:Review: Active/ })).toBeInTheDocument();
act(() => socketHandlers.get('cos:tasks:changed')({ task: {
id: 'preflight-demand-1', status: 'completed',
@@ -401,13 +401,13 @@ describe('PullRequestsTab', () => {
preflightResultTaskId: 'app-improve-17',
preflight: { phase: 'done', outcome: 'handed-off', steps: [] } },
} }));
- expect(await screen.findByRole('link', { name: /PR review: Active/ })).toBeInTheDocument();
+ expect(await screen.findByRole('link', { name: /Untrusted:Review: Active/ })).toBeInTheDocument();
act(() => socketHandlers.get('cos:tasks:changed')({ task: {
id: 'app-improve-17', status: 'completed',
metadata: { app: 'app-1', analysisType: 'pr-reviewer', targetPullRequest: 17 },
} }));
- expect(await screen.findByRole('link', { name: /PR review: Completed/ }))
+ expect(await screen.findByRole('link', { name: /Untrusted:Review: Completed/ }))
.toHaveAttribute('href', '/cos/tasks?task=app-improve-17&source=internal');
});
@@ -418,7 +418,7 @@ describe('PullRequestsTab', () => {
await renderTab();
expect(await screen.findByText('Fix the save path')).toBeInTheDocument();
- expect(screen.queryByRole('button', { name: 'PR review' })).not.toBeInTheDocument();
+ expect(screen.queryByRole('button', { name: 'Untrusted:Review' })).not.toBeInTheDocument();
fireEvent.click(screen.getByRole('button', { name: 'Do:Review' }));
@@ -467,7 +467,7 @@ describe('PullRequestsTab', () => {
expect(await screen.findByRole('alert')).toHaveTextContent('Resolve & merge failed');
expect(screen.queryByRole('link', { name: 'Abuse Guard setup' })).not.toBeInTheDocument();
- expect(screen.queryByRole('button', { name: 'Retry PR review' })).not.toBeInTheDocument();
+ expect(screen.queryByRole('button', { name: 'Retry Untrusted:Review' })).not.toBeInTheDocument();
fireEvent.click(screen.getByRole('button', { name: 'Retry Resolve & merge' }));
await waitFor(() => expect(api.resolveAppPullRequest).toHaveBeenCalledWith(
'app-1', 17, { provider: undefined, model: undefined, effort: undefined, reviewMode: 'delegated' },
@@ -484,8 +484,8 @@ describe('PullRequestsTab', () => {
await renderTab();
- expect(await screen.findByRole('link', { name: /PR review: Active/ })).toBeInTheDocument();
- expect(screen.queryByRole('button', { name: /PR review/ })).not.toBeInTheDocument();
+ expect(await screen.findByRole('link', { name: /Untrusted:Review: Active/ })).toBeInTheDocument();
+ expect(screen.queryByRole('button', { name: /Untrusted:Review/ })).not.toBeInTheDocument();
});
it('omits the pr-reviewer action on a row the server marked ineligible', async () => {
@@ -494,7 +494,7 @@ describe('PullRequestsTab', () => {
await renderTab();
expect(await screen.findByRole('button', { name: /Resolve & merge/ })).toBeInTheDocument();
- expect(screen.queryByRole('button', { name: /PR review/ })).not.toBeInTheDocument();
+ expect(screen.queryByRole('button', { name: /Untrusted:Review/ })).not.toBeInTheDocument();
});
it('merges a request directly once the inline confirm is answered', async () => {
diff --git a/client/src/hooks/useOnDemandTaskToast.js b/client/src/hooks/useOnDemandTaskToast.js
index ed8a3aa9c7..6f4fa4ed8c 100644
--- a/client/src/hooks/useOnDemandTaskToast.js
+++ b/client/src/hooks/useOnDemandTaskToast.js
@@ -39,7 +39,7 @@ const PR_REVIEWER_REASON_LABELS = {
'no-external-open-prs': 'no open external pull requests to review',
'target-pull-request-not-reviewable': "that pull request isn't eligible right now (not open against the default branch, or authored by a trusted collaborator)",
'security-scan-report-pending': 'a security scan for this pull request is already in progress',
- 'security-guard-process-failed': 'Prompt Guard stopped before an agent started. Open Models → LLMs → Abuse Guard to check or repair setup, then retry PR review. The failure record is in CoS → Tasks',
+ 'security-guard-process-failed': 'Prompt Guard stopped before an agent started. Open Models → LLMs → Abuse Guard to check or repair setup, then retry Untrusted:Review. The failure record is in CoS → Tasks',
'security-guard-not-ready': "the local model-abuse classifier (Settings → Models → LLMs → Abuse Guard) isn't ready — finish or repair its setup, then try again",
};
diff --git a/docs/features/messages-security.md b/docs/features/messages-security.md
index eb1959657d..8540037fab 100644
--- a/docs/features/messages-security.md
+++ b/docs/features/messages-security.md
@@ -51,7 +51,7 @@ PR review does not execute contributor tests or apply patches in its default sta
PR review stage provider/model pins are strict: disabled, unsupported or unavailable selections stop the review rather than switching to a subscription provider or a different model. Local classifier setup/runtime failures and review provider/model failures raise a notification and queue a deduplicated CoS investigation. Investigation prompts contain only server-owned diagnostics, never contributor text, model output or transport errors. A malicious-content finding is not an infrastructure incident.
-The PR page follows the saved stage providers by default. Its **Use Run with for PR review eligibility** checkbox explicitly overrides only the eligibility stage for that invocation; the final review keeps its own stage settings. Stage-specific settings take precedence over broader schedule/app defaults. The shared Abuse Guard text API selection does not replace these PR stage selections: Stage 1 always uses the dedicated local classifier, while stages 2 and 3 use the providers configured in **CoS > Schedule > PR Reviewer**.
+The PR page follows the saved stage providers by default. Its **Also run Untrusted:Review on the Run with provider** checkbox explicitly overrides only the eligibility stage for that invocation; the final review keeps its own stage settings. Stage-specific settings take precedence over broader schedule/app defaults. The shared Abuse Guard text API selection does not replace these PR stage selections: Stage 1 always uses the dedicated local classifier, while stages 2 and 3 use the providers configured in **CoS > Schedule > PR Reviewer**.
After an approval, the deterministic coordinator merges a green, mergeable PR or requests GitHub auto-merge while CI is pending. The enable-only GraphQL mutation names the reviewed head SHA, so it cannot authorize a different commit or immediately bypass pending checks on an unprotected repository. A rejected auto-merge request raises a notification and retains the approval for the existing bounded merge poller.