auto-mode can loop repeated startup notifications and the same select prompt without dispatching work

[AriaShishegaran]

Summary

/gsd auto can get stuck in a startup/UI re-entry loop where it repeatedly emits the same notify/select events and never dispatches a real unit.

What I observed

In headless mode, /gsd auto produced output like:

• repeated notify: Debug logging enabled -> ...
• repeated select: GSD — M068: M068
• no actual tool dispatches
• timeout with no real work started

This happened before any substantive unit execution, so it looks like a startup / guided-flow re-entry loop rather than a task-level failure.

Expected

If auto-mode determines the next actionable milestone/slice/task, it should proceed into dispatch.
If it needs the guided flow instead, it should present one prompt once, not the same prompt repeatedly.

Actual

The same startup notification and the same select prompt repeat until timeout.

Why this matters

This makes /gsd auto look alive while doing no work.
It is especially frustrating in milestone recovery scenarios because the operator sees auto-mode looping instead of either:

• continuing real work
• or stopping with one clear blocking prompt

Notes that may help debug

This looks related to startup / command re-entry around:

• src/resources/extensions/gsd/commands/handlers/auto.ts
• src/resources/extensions/gsd/auto.ts
• src/resources/extensions/gsd/auto-start.ts
• src/resources/extensions/gsd/guided-flow.ts

The repeated Debug logging enabled notifications suggest startup is re-entering more than once inside a single command lifecycle.

[Tibsfox]

This issue is in the auto-mode dispatch/stuck-loop cluster. Several PRs from the recent fix batch address related root causes:

• fix(gsd): cap run-uat dispatch attempts to prevent infinite replay loop #3686 — run-uat replay cap (prevents infinite dispatch loops)
• fix(gsd): force re-validation when verdict is needs-remediation #3670 — needs-remediation revalidation (breaks completing-milestone deadlocks)
• fix(gsd): auto-remediate stale slice DB status when SUMMARY exists #3673 — auto-remediate stale slice status (breaks complete-slice re-dispatch loops)
• fix(gsd): set slice sequence at all three insertion sites #3697 — slice sequence at insertion sites (prevents positional deadlocks)
• fix(gsd): prompt ordering, correct tools, crash guard, compaction fix #3696 — uncaught exception handler + compaction fix

If this specific issue persists after merging, the remaining root cause should be narrower and easier to isolate.

[jeremymcs]

Hey — we now use /gsd-forensics to help diagnose bug reports. Could you run /gsd-forensics and paste the output here? If you're on v2.7x and still experiencing this, the forensics output will help us triage much faster. Thanks!

[trek-e]

Closing as stale — maintainers requested additional diagnostic information 4+ days ago with no response. Please reopen and provide the requested output if you're still experiencing this.