qwen-code proposal: one loop-protection setting (off/warn/stop) with a reminder before stopping
qqqys · ghdev · 2026-10-20
A feature request on QwenLM/qwen-code proposes unifying tool-call loop detection under a single setting, model.loopProtection: off / warn / stop, honored across the interactive TUI, JSON modes, subagents, and daemon/ACP sessions.
Key points:
- The default stop mode keeps every existing stop point but adds one corrective reminder from each always-on detector before halting; warn only warns, off disables stopping entirely (telemetry still logs detections), with a 1000-call backstop ending stuck turns.
- Always-on detectors include consecutive identical tool calls (stop at 5th, warn at 3rd), shell inspection stagnation (8th repeated git status/diff), invalid tool params stagnation, and an adaptive cap past 100 calls with one call repeated 6 times.
- The issue is labeled priority/P2 and need-discussion, with 3 comments so far.
More from coding & agent
- Hallmark v2 questioned: how does it check taste without a human looking? — petrusenko_max · 2026-10-23
- Laude Institute launches FrontierCS 2: agents must beat deleted paper contributions blind — lateinteraction · 2026-10-23
- Index Network open-sources Social Discovery Protocol: agents negotiate fit on your behalf — carsonfarmer · 2026-10-23
- Installed a formal-verification skill: it silently installed Java, MacBook runs 10 degrees hotter — zealcaiden · 2026-10-23
- OPAL: an LLM agent rewrites the optimizer's source code during training — RobertTLange · 2026-10-23
- OpenAI's Decisions API hits GA: near real-time routing up to 10x faster than GPT-6 Luna — OpenAIDevs · 2026-10-23