feat(e2e): let the e2e harness drive the warehouse flow

e2e-harness/ARCHITECTURE.md

@@ -21,7 +21,8 @@ e2e-harness/
   wizard-ci-driver.ts   WizardCiDriver — read_state / perform_action over the store
   action-registry.ts    screen → the actions legal on it (+ NO_ACTION_SCREENS)
   e2e-profile.ts        WizardE2eProfile + decideE2eAction — the scripted walk policy
-  profiles.ts           per-program profiles + profileFor(programId)
+  profiles.ts           per-program profiles + profileFor(programId) + resolveE2eProfile(env)
+  e2e-result.ts         the E2E_RESULT_JSON payload: E2eRunRecorder + buildE2eResult
   tui-capture.ts        run a command in a PTY (node-pty) + read its real screen (@xterm/headless)
 scripts/
   tui-host.no-jest.ts   the real-TUI host: startTUI + WizardCiDriver, MODE=fixed | serve
@@ -78,6 +79,48 @@ the program's entry (typed by `WizardE2eProfile`); the host asks
 `decideE2eAction(state, profile)` what to commit on each screen. The (screen →
 decision) trace is snapshot-tested offline in `__tests__/` (`jest -u` to update).
 
+## Driving the agent-in-the-loop layer
+
+Two decision points ask a person to act, and the harness stands in for them.
+
+`wizard_ask` overlay. A `ci` session normally has no ask bridge at all. The host
+sets `session.e2eAsk` from `E2E_ASK=true`, which keeps the bridge wired (see
+`shouldDisableAsk`). The profile then answers every question in the batch:
+`askAnswers` routes a question to a value, else the first option, else the `'e2e'`
+sentinel. Route credentials with `${ENV_VAR}` values, never literals.
+
+Mark a credential rule `"secret": true`. The skill names its own questions, so a
+rule matches text the agent controls — without the flag, a question called
+`password` takes the value, and by omitting `sensitive: true` gets it back
+unvaulted in plaintext. A secret rule answers only free text flagged sensitive
+and refuses every other question shape; refusals land in the payload's
+`refusedIds` / `refusedAsks`, so a run that withholds a credential says so
+instead of looking like an ordinary sentinel.
+
+Task-notice overlay. `profile.notice` decides `keep` or `decline`. `E2E_NOTICE`
+overrides it per run.
+
+Both env inputs are folded into the profile by `resolveE2eProfile`, at load.
+`decideE2eAction` must stay pure — it reads no env and no store.
+
+## The result payload
+
+A `MODE=fixed` run writes `E2E_RESULT_JSON`: the run phase and screens walked,
+plus every ask batch, every task notice, the task list, the detected warehouse
+sources, the program's report file, and the abort reason. `e2e-result.ts` builds
+it.
+
+**Only question prompt text and question ids go in the payload. No answer value
+ever does.** The decision function reports ids and a keep/decline verdict, so the
+recorder never holds an answer. Keep it that way — the workbench scans the file
+for its own injected credentials.
+
+The report file is the one payload field whose *content* comes from the app
+directory, which is a checkout the run does not control. `readReportFile` reads
+only a regular file inside `appDir` — never a symlink, and never through a
+symlinked parent — so a committed `posthog-warehouse-report.md` pointing at a
+host file cannot copy it into the payload.
+
 ## Visual-regression snapshots (the workbench flow)
 
 [wizard-workbench](https://github.com/PostHog/wizard-workbench) runs the CI route

e2e-harness/__tests__/__snapshots__/e2e-flow-snapshot.test.ts.snap

@@ -86,6 +86,7 @@ exports[`e2e flow snapshot — posthog-integration > Next.js (with a setup quest
     "ask": "first",
     "healthCheck": "dismiss",
     "mcp": "skip",
+    "notice": "keep",
     "setup": "first",
     "skills": "delete",
     "slack": "skip",
@@ -171,3 +172,76 @@ exports[`e2e flow snapshot — posthog-integration > Node (no setup question) wa
   ],
 }
 `;
+
+exports[`e2e flow snapshot — warehouse-source > walks intro → auth → run → outro → skills 1`] = `
+{
+  "profile": {
+    "ask": "first",
+    "askAnswers": [
+      {
+        "match": "prefix",
+        "value": "\${E2E_SOURCE_PREFIX}",
+      },
+      {
+        "match": "stripe",
+        "secret": true,
+        "value": "\${E2E_STRIPE_API_KEY}",
+      },
+      {
+        "match": "host|hostname|server|endpoint",
+        "value": "\${E2E_PG_HOST}",
+      },
+      {
+        "match": "port",
+        "value": "\${E2E_PG_PORT}",
+      },
+      {
+        "match": "database|db_?name",
+        "value": "\${E2E_PG_DATABASE}",
+      },
+      {
+        "match": "password|passwd|pwd",
+        "secret": true,
+        "value": "\${E2E_PG_PASSWORD}",
+      },
+      {
+        "match": "user|username|role",
+        "value": "\${E2E_PG_USER}",
+      },
+      {
+        "match": "schema",
+        "value": "public",
+      },
+    ],
+    "healthCheck": "dismiss",
+    "mcp": "skip",
+    "notice": "keep",
+    "setup": "first",
+    "skills": "delete",
+    "slack": "skip",
+  },
+  "program": "warehouse-source",
+  "trace": [
+    {
+      "action": "confirm_setup",
+      "screen": "warehouse-intro",
+    },
+    {
+      "action": "(external)",
+      "screen": "auth",
+    },
+    {
+      "action": "(external)",
+      "screen": "run",
+    },
+    {
+      "action": "dismiss_outro",
+      "screen": "outro",
+    },
+    {
+      "action": "keep_skills",
+      "screen": "keep-skills",
+    },
+  ],
+}
+`;

e2e-harness/__tests__/e2e-flow-snapshot.test.ts

@@ -129,3 +129,26 @@ describe('e2e flow snapshot — metrics', () => {
     expect(trace.length).toBeLessThan(40);
   });
 });
+
+describe('e2e flow snapshot — warehouse-source', () => {
+  it('walks intro → auth → run → outro → skills', () => {
+    expect({
+      program: 'warehouse-source',
+      profile: profileFor(Program.WarehouseSource),
+      trace: traceFlow(Integration.javascriptNode, Program.WarehouseSource),
+    }).toMatchSnapshot();
+  });
+
+  it('reaches a terminal decision instead of stalling on the intro', () => {
+    const trace = traceFlow(
+      Integration.javascriptNode,
+      Program.WarehouseSource,
+    );
+    expect(trace[0]).toEqual({
+      screen: 'warehouse-intro',
+      action: 'confirm_setup',
+    });
+    expect(trace.at(-1)?.action).toBe('keep_skills');
+    expect(trace.length).toBeLessThan(40);
+  });
+});