If you're in early onboarding (often Module 2) and your agent keeps returning an error instead of a normal reply, don't keep random retrying. Use this 4-minute flow.

Step 1) Capture One Clean Failure

Step 2) Rule Out Surface/Context Confusion

• Retry the same tiny prompt in one other surface (for example, dashboard chat vs Discord).
• If one surface works and another fails, this is usually context/routing/policy, not model quality.
• If both fail with the same error, continue to runtime-health checks.

Step 3) Check for Runtime Health Pattern

If your errors mention things like runtime not healthy, health 502, or Gateway tool invoke failed: 500, you're likely dealing with runtime-side instability, not prompt quality.

Step 4) Open the Ticket the Right Way

Use the active support path (if legacy #help is moved, use heyron.ai/support). Paste this packet:

Issue summary: Module 2 setup, agent errors on every reply
Exact error text:
Where it failed (dashboard / Discord DM / server):
Tiny canary prompt used:
Result in each surface tested:
Timestamp + timezone:
Any recent reconnect/restart attempt + outcome:

High-Friction Cases

“How do I create a help ticket?”

Use: https://heyron.ai/support. Include exact error text and one screenshot. Don’t post only “it doesn’t work.”

Community says “start over” — should I?

Only after you capture one clean failure and one tiny canary test. Starting over blindly can erase useful evidence and slow support.

The error happens in every reply

That usually points to runtime/config health, not bad prompting. Use the packet above and escalate once with full evidence.

“I reconnected/re-entered email and now it’s worse”

Do one clean reconnect flow only, then rerun the tiny canary prompt. Repeating reconnects can create noisy state and make diagnosis harder.

Which error matters most: 500/502 vs token mismatch?

If you see both in the same window, prioritize runtime health errors (health 502, 500) first. If runtime is healthy but auth still fails, then follow Gateway Token Mismatch.

I opened a ticket but got silence

Don’t open duplicates. Reply in the same ticket with one fresh canary result, exact current error text, and timestamp/timezone. This speeds triage and avoids fragmented threads.