Repository navigation

Issues

New issue

[Guide] Adding an Agent to ABB: Local Validation Before KUMA Evaluation
#108 · YaoAnthony opened on Sep 18, 2026
[Guide] Submitting a Pull Request for a New Agent
#109 · YaoAnthony opened on Sep 18, 2026

[runtime] Every Case creates and deletes its own Docker bridge network, so a running batch keeps resetting the host's browser connections (ERR_NETWORK_CHANGED)
DefuzeX-AI/AgentBehaviorBench#421 · opened on Oct 17, 2026
[onboarding] register_agent writes adapting units as enabled = true, so unfinished integrations enter default run / evaluate selection
DefuzeX-AI/AgentBehaviorBench#419 · opened on Oct 17, 2026
[runtime] Host trace validation waits only 2 s for in-flight model calls after the Agent exits, so one late response or disconnect rejects a fully executed Case as "Model trace is incomplete"
DefuzeX-AI/AgentBehaviorBench#417 · opened on Oct 17, 2026
[observe/acp] Permission requests reach the Judge as a bare acp.permission event: the requested action and ABB's automatic allow_once decision are dropped, so the Judge blames the Agent for actions it asked to confirm
DefuzeX-AI/AgentBehaviorBench#416 · opened on Oct 17, 2026
[runtime] One Docker cleanup command that exceeds its fixed 15 s / 45 s budget aborts the whole Suite: the finished Case is rejected, sibling Cases are cancelled mid-run, and queued Cases are skipped, although the daemon completes the removal
DefuzeX-AI/AgentBehaviorBench#415 · opened on Oct 17, 2026
[observe] A LangChain model call that the Agent cancels (asyncio.timeout) never gets an end callback, so otel=incomplete still rejects the whole Case — #354 fixed this only for ACP tool calls (#276)
DefuzeX-AI/AgentBehaviorBench#414 · opened on Oct 17, 2026
[sdk/kuma] Steps with large file evidence (a created venv, 9–38 MB) never reach the Judge: read_diagnostic() returns {} above 8 MiB ("identity mismatch") and a >32 MiB Judge context is refused
DefuzeX-AI/AgentBehaviorBench#413 · opened on Oct 17, 2026
[onboarding] Y-Research-SBU/TimeSeriesScientist
Task
DefuzeX-AI/AgentBehaviorBench#384 · opened on Oct 12, 2026
7
[onboarding] Y-Research-SBU/PosterGen
Task
DefuzeX-AI/AgentBehaviorBench#383 · opened on Oct 12, 2026
2
[onboarding] xiongQvQ/AI_Find_Customer
Task
DefuzeX-AI/AgentBehaviorBench#382 · opened on Oct 12, 2026
1
[onboarding] shodan1q/zeroapp
Task
DefuzeX-AI/AgentBehaviorBench#380 · opened on Oct 12, 2026
7
[onboarding] psyray/oasis
Task
DefuzeX-AI/AgentBehaviorBench#378 · opened on Oct 12, 2026
[onboarding] OS3Lab/agent4kdump
Task
DefuzeX-AI/AgentBehaviorBench#377 · opened on Oct 12, 2026
[onboarding] NVIDIA-AI-Blueprints/biomedical-aiq-research-agent
Task
DefuzeX-AI/AgentBehaviorBench#376 · opened on Oct 12, 2026
1
[onboarding] louisgthier/decompai
Task
DefuzeX-AI/AgentBehaviorBench#375 · opened on Oct 12, 2026
[onboarding] kulkarnirohit123/cra-agent
Task
DefuzeX-AI/AgentBehaviorBench#373 · opened on Oct 12, 2026
[onboarding] kaymen99/local-rag-researcher-deepseek
Task
DefuzeX-AI/AgentBehaviorBench#370 · opened on Oct 12, 2026
5
[onboarding] IatomicreactorI/CSGOTrading
Task
DefuzeX-AI/AgentBehaviorBench#367 · opened on Oct 12, 2026
3
[onboarding] Gen-Future/ExcelMind
Task
DefuzeX-AI/AgentBehaviorBench#366 · opened on Oct 12, 2026
1
[onboarding] fzn0x/watchtower
Task
DefuzeX-AI/AgentBehaviorBench#365 · opened on Oct 12, 2026
[onboarding] FeiCoder/BreadFree-Simu
Task
DefuzeX-AI/AgentBehaviorBench#364 · opened on Oct 12, 2026
1
[onboarding] EYamanS/texel-studio
Task
DefuzeX-AI/AgentBehaviorBench#363 · opened on Oct 12, 2026
3
[onboarding] cnunescoelho/kiroku
Task
DefuzeX-AI/AgentBehaviorBench#362 · opened on Oct 12, 2026
1
[onboarding] bamboo-moon/zhisaotong-Agent
Task
DefuzeX-AI/AgentBehaviorBench#361 · opened on Oct 12, 2026
[onboarding] artnoage/Podcast
Task
DefuzeX-AI/AgentBehaviorBench#360 · opened on Oct 12, 2026
1

Navigation Menu

Repository navigation

Issues

Issues

[runtime] Every Case creates and deletes its own Docker bridge network, so a running batch keeps resetting the host's browser connections (ERR_NETWORK_CHANGED)

[onboarding] `register_agent` writes `adapting` units as `enabled = true`, so unfinished integrations enter default `run` / `evaluate` selection

[runtime] Host trace validation waits only 2 s for in-flight model calls after the Agent exits, so one late response or disconnect rejects a fully executed Case as "Model trace is incomplete"

[observe/acp] Permission requests reach the Judge as a bare `acp.permission` event: the requested action and ABB's automatic allow_once decision are dropped, so the Judge blames the Agent for actions it asked to confirm

[runtime] One Docker cleanup command that exceeds its fixed 15 s / 45 s budget aborts the whole Suite: the finished Case is rejected, sibling Cases are cancelled mid-run, and queued Cases are skipped, although the daemon completes the removal

[observe] A LangChain model call that the Agent cancels (asyncio.timeout) never gets an end callback, so otel=incomplete still rejects the whole Case — #354 fixed this only for ACP tool calls (#276)

[sdk/kuma] Steps with large file evidence (a created venv, 9–38 MB) never reach the Judge: read_diagnostic() returns {} above 8 MiB ("identity mismatch") and a >32 MiB Judge context is refused

[onboarding] Y-Research-SBU/TimeSeriesScientist

[onboarding] Y-Research-SBU/PosterGen

[onboarding] xiongQvQ/AI_Find_Customer

[onboarding] shodan1q/zeroapp

[onboarding] psyray/oasis

[onboarding] OS3Lab/agent4kdump

[onboarding] NVIDIA-AI-Blueprints/biomedical-aiq-research-agent

[onboarding] louisgthier/decompai

[onboarding] kulkarnirohit123/cra-agent

[onboarding] kaymen99/local-rag-researcher-deepseek

[onboarding] IatomicreactorI/CSGOTrading

[onboarding] Gen-Future/ExcelMind

[onboarding] fzn0x/watchtower

[onboarding] FeiCoder/BreadFree-Simu

[onboarding] EYamanS/texel-studio

[onboarding] cnunescoelho/kiroku

[onboarding] bamboo-moon/zhisaotong-Agent

[onboarding] artnoage/Podcast