Agent in Autopilot repeatedly ran tasklist and appended output to context, consuming about 936 credits in 17 minutes.
- What was harmed
- Money: usage, API bills, ad spend
- Severity
- Moderate: real damage that the user could repair, or money spent by mistake
- Could have been
- Severe: unbounded loop in Autopilot could burn far more credits
- Money lost
- $20
- Recovered
- The report does not say
- Reported
- 5 Nov 2026
Read the original report (Kiro #12303 on GitHub)
How sure are we?
Further right means the case carries more weight.
Evidence
CBA
A: The vendor engaged with the report.
Who wrote the report
The agentUser, quoting agentThe user
Written by the user
Was anyone watching?
Turned downNot statedOversight on
Oversight reduced: Autopilot enabled
- Vendor response
- Maintainer confirmed looping behavior, marked as bug
- Other trackers
- Not in Manifold, Eon, AIR or METR