Bug: expect-cli 0.0.24 and 0.0.25 stall with 'Agent produced no output for 180s' on localhost app

[caseyjkey]

Summary

expect-cli stalls for 180 seconds and then exits with Streaming failed: Agent produced no output for 180s — the agent may be stalled.

I reproduced this on both 0.0.24 and 0.0.25 while testing a localhost app.

Environment

• expect-cli --version: 0.0.24, then upgraded to 0.0.25
• Node: /home/trill/.nvm/versions/node/v23.6.0/bin/node
• OS: Linux
• Target app URL: http://127.0.0.1:8080

Reproduction

Tried these commands:

expect-cli -u http://127.0.0.1:8080 -m "Start from the login screen. Test login affordances, then use Start with Demo Mode..." -y
expect-cli -u http://127.0.0.1:8080 -m "Start from the login screen and enter Demo Mode. Exercise the item goal workflow end to end..." -y
expect-cli -u http://127.0.0.1:8080 -m "Start from the login screen and use Start with Demo Mode for Stripe-related features..." -y
expect-cli -u http://127.0.0.1:8080 -m "Open the app, use Start with Demo Mode, and do a brief smoke test of the main screen." -y --verbose

All of them stalled after about 3 minutes with the same error.

What I ruled out

• The localhost app was reachable from the machine
• expect-cli was upgraded from 0.0.24 to 0.0.25
• A local Chrome instance was launched with CDP enabled:

/usr/bin/google-chrome-stable \
  --headless=new \
  --disable-gpu \
  --no-sandbox \
  --remote-debugging-port=9222 \
  --remote-debugging-address=127.0.0.1 \
  --user-data-dir=/tmp/expect-chrome-profile \
  --no-first-run \
  --no-default-browser-check \
  about:blank

That Chrome process successfully reported:

DevTools listening on ws://127.0.0.1:9222/devtools/browser/...

So browser discovery was available, but expect-cli still timed out waiting for agent output.

Actual result

After ~180s:

Streaming failed: Agent produced no output for 180s — the agent may be stalled

The verbose trace also included the CDP auto-discovery code path in the stack, but once Chrome was manually running with --remote-debugging-port=9222, the final failure remained the same 180s agent stall.

Expected result

expect-cli should either:

• proceed with the browser test, or
• surface a more specific actionable error if the underlying agent / browser session cannot start

Notes

• .expect/logs.md remained empty in this project during reproduction
• 0.0.25 still does not accept --cookies; help only shows default cookie extraction and --no-cookies

[aidenybai]

Improved the error message in first-minor branch. The 180s stall timeout now shows actionable troubleshooting: verify your dev server is running, use -u to set the correct URL, and run with --verbose for more details.

[flosrn]

Reproduced the same failure shape on expect-cli 0.1.3 while testing a localhost Next.js app via Codex ACP. Command shape: expect-cli tui -u http://localhost:3004 -m "..." --no-cookies -y. Local versions: expect-cli 0.1.3, Node v25.2.1, macOS arm64. The app was reachable and the dev server was healthy; the run created an ACP session, then exited after ~180s with: Streaming failed: Agent produced no output for 180s. .expect/logs.md only showed ACP initialization. I also found a stale orphaned /opt/homebrew/lib/node_modules/expect-cli/dist/browser-mcp.js process afterward, matching the cleanup issue described in #98.