[Feat] Agents screen their own diff for risky hunks before they ship

apps/api/src/handlers/tasks/getDiffRiskHints.ts

@@ -0,0 +1,76 @@
+import type { Context } from 'hono';
+import { z } from 'zod';
+
+import {
+  DIFF_RISK_HINTS_MAX_DIFF_CHARS,
+  screenDiffRiskHints,
+} from '@roomote/cloud-agents/server';
+import { db, eq, tasks, taskRuns } from '@roomote/db/server';
+
+import type { Variables } from '../../types';
+import type { McpAuth } from '../mcp/middleware';
+import { isRunTokenContext } from '../mcp/proxy-utils';
+import { logHandlerError } from '../utils';
+
+const bodySchema = z.object({
+  diff: z.string().min(1).max(DIFF_RISK_HINTS_MAX_DIFF_CHARS),
+});
+
+/**
+ * Risk hints for a task's own diff: the pull request review pre-screen, run
+ * before the agent ships. The sandbox sends the diff; the judgment model key
+ * stays here.
+ */
+export async function getDiffRiskHints(
+  c: Context<{ Variables: Variables & { mcpAuth: McpAuth } }>,
+): Promise<Response> {
+  const auth = c.get('mcpAuth').authContext;
+
+  if (!isRunTokenContext(auth)) {
+    return c.json({ error: 'Diff risk hints require a task run token' }, 403);
+  }
+
+  const runId = Number(c.req.param('runId'));
+
+  if (!Number.isInteger(runId) || runId <= 0) {
+    return c.json({ error: 'Invalid task run id' }, 400);
+  }
+
+  if (auth.runId !== runId) {
+    return c.json(
+      { error: 'Task run token does not match requested task run' },
+      403,
+    );
+  }
+
+  const parsed = bodySchema.safeParse(await c.req.json().catch(() => null));
+
+  if (!parsed.success) {
+    return c.json(
+      {
+        error: `Invalid request: send the diff as \`diff\` (at most ${DIFF_RISK_HINTS_MAX_DIFF_CHARS} characters).`,
+      },
+      400,
+    );
+  }
+
+  try {
+    const [row] = await db
+      .select({ title: tasks.title })
+      .from(taskRuns)
+      .innerJoin(tasks, eq(tasks.id, taskRuns.taskId))
+      .where(eq(taskRuns.id, runId))
+      .limit(1);
+
+    if (!row) {
+      return c.json({ error: 'Task run not found' }, 404);
+    }
+
+    return c.json(
+      await screenDiffRiskHints({ title: row.title, diff: parsed.data.diff }),
+    );
+  } catch (error) {
+    logHandlerError('getDiffRiskHints', error);
+    return c.json({ error: 'Failed to screen the diff' }, 500);
+  }
+}

apps/api/src/handlers/tasks/index.ts

@@ -19,6 +19,7 @@ import { manageSourceControl } from './manageSourceControl';
 import { updateTaskModelSelection } from './updateModelSelection';
 import { listTaskModels } from './listModels';
 import { saveTaskMemory } from './saveTaskMemory';
+import { getDiffRiskHints } from './getDiffRiskHints';
 import { recordAutomationResult } from './recordAutomationResult';
 import { updatePersonalization } from './updatePersonalization';
 
@@ -43,4 +44,5 @@ tasksRouter.post('/:taskId/task_suggestions', submitTaskSuggestions);
 tasksRouter.post('/:taskId/automation_result', recordAutomationResult);
 tasksRouter.post('/:taskId/mcp_recommendations', submitMcpRecommendations);
 tasksRouter.post('/runs/:runId/memory', saveTaskMemory);
+tasksRouter.post('/runs/:runId/diff_risk_hints', getDiffRiskHints);
 tasksRouter.post('/runs/:runId/personalization', updatePersonalization);

apps/docs/models.mdx

@@ -355,6 +355,9 @@ When a judgment model is on, Roomote asks it first for:
 - choosing the forum tag when Roomote opens a thread in a Discord forum channel
 - whether a settled session or task turn holds something worth saving to
   [Memory](/memory) that the agent did not record itself
+- ranking a task's own changed hunks by defect risk during its self-review,
+  before it pushes or opens a pull request (the same pre-screen pull request
+  reviews use), as advisory hints the agent re-reads
 
 Roomote acts on the judgment model only when it is confident. When it is
 unsure, unavailable, or returns an error, Roomote keeps the behavior it has