[Bug/UX] ShellAgent hangs indefinitely during docker-compose execution without timeout or stderr feedback

[Adhders]

2026-04-09 17:19:31.594 [INFO ] tool_layer.py:295 - OpenSpace ready to use!
2026-04-09 17:19:31.594 [INFO ] tool_layer.py:296 - ============================================================
2026-04-09 17:19:50.809 [INFO ] ui_integration.py:79 - UI monitoring started
2026-04-09 17:19:50.809 [INFO ] tool_layer.py:345 - ============================================================
2026-04-09 17:19:50.809 [INFO ] tool_layer.py:346 - Task: openspace --query "Create a monitoring dashboard for my Docker containers，with Real-time status disp...
2026-04-09 17:19:50.809 [INFO ] tool_layer.py:347 - ============================================================
2026-04-09 17:19:50.809 [INFO ] tool_layer.py:357 - Task ID: task_e3d49f500184
2026-04-09 17:19:50.820 [INFO ] manager.py:534 - Recording started: logs/recordings/task_e3d49f500184_20260409_171950
2026-04-09 17:19:50.820 [INFO ] tool_layer.py:378 - Recording started: task_e3d49f500184
2026-04-09 17:19:50.820 [INFO ] tool_layer.py:388 - Workspace: logs/recordings/task_e3d49f500184_20260409_171950
2026-04-09 17:19:58.540 [WARNING ] skill_ranker.py:351 - Skill embedding generation failed after 3 attempts: HTTP Error 403: Forbidden
2026-04-09 17:19:58.541 [INFO ] registry.py:540 - Skill pre-filter: 38 → 15 candidates (BM25+embedding, threshold=10)
2026-04-09 17:19:58.548 [INFO ] utils.py:3995 -
LiteLLM completion() model= doubao-seed-2.0-pro; provider = openai
2026-04-09 17:20:06.720 [INFO ] registry.py:731 - Skill selection plan: Deploy a cAdvisor-Prometheus-Grafana monitoring stack to collect real-time Docker container status, CPU, memory, disk I/O and network I/O metrics, then provision a dedicated Grafana dashboard to visualize all required monitoring data.
2026-04-09 17:20:06.720 [INFO ] registry.py:489 - LLM skill selection: [docker-monitoring-cadvisor-prometheus-grafana__v1_aeb5f200]
2026-04-09 17:20:06.721 [INFO ] manager.py:171 - Recorded skill selection: 1 selected via llm (from 15 available)
2026-04-09 17:20:06.721 [INFO ] grounding_agent.py:102 - Skill context set: docker-monitoring-cadvisor-prometheus-grafana__v1_aeb5f200
2026-04-09 17:20:06.721 [INFO ] tool_layer.py:724 - Injected 1 active skill(s): ['docker-monitoring-cadvisor-prometheus-grafana__v1_aeb5f200']
2026-04-09 17:20:06.721 [INFO ] tool_layer.py:426 - [Phase 1 — Skill] Executing with skill guidance (max 30 iterations)...
2026-04-09 17:20:06.721 [INFO ] grounding_agent.py:211 - Grounding Agent: Processing instruction at step 0
2026-04-09 17:20:06.721 [INFO ] grounding_agent.py:577 - GroundingAgent selected 9 tools (auto-search) from 3 backends [skill-augmented]
2026-04-09 17:20:06.722 [INFO ] grounding_agent.py:601 - Added retrieve_skill tool for mid-iteration skill retrieval
2026-04-09 17:20:06.730 [INFO ] manager.py:143 - Recorded 10 retrieved tools (with search debug info: False)
2026-04-09 17:20:06.730 [INFO ] grounding_agent.py:533 - Injected active skill context (1 skill(s))
2026-04-09 17:20:06.731 [INFO ] grounding_agent.py:283 - Grounding Agent: Iteration 1/30
2026-04-09 17:20:06.731 [INFO ] client.py:166 - [SCHEMA_SANITIZE] Prepared 10 tools for LLM (from 10 total)
2026-04-09 17:20:06.735 [INFO ] utils.py:3995 -
LiteLLM completion() model= doubao-seed-2.0-pro; provider = openai
2026-04-09 17:20:12.456 [INFO ] client.py:703 - Executing 1 tool calls...
2026-04-09 17:20:12.456 [INFO ] client.py:753 - Calling run_shell with args: {"command": "docker compose version"}
2026-04-09 17:20:12.610 [INFO ] client.py:814 - Tool execution completed, 1 tools executed
2026-04-09 17:20:12.611 [INFO ] grounding_agent.py:349 - Iteration 1 - Has tool calls: True, Tool results: 1, Content length: 0 chars
2026-04-09 17:20:12.611 [INFO ] grounding_agent.py:283 - Grounding Agent: Iteration 2/30
2026-04-09 17:20:12.611 [INFO ] grounding_agent.py:294 - Skill context removed from messages after first iteration
2026-04-09 17:20:12.611 [INFO ] client.py:166 - [SCHEMA_SANITIZE] Prepared 10 tools for LLM (from 10 total)
2026-04-09 17:20:12.612 [INFO ] utils.py:3995 -
LiteLLM completion() model= doubao-seed-2.0-pro; provider = openai
2026-04-09 17:20:36.937 [INFO ] client.py:703 - Executing 1 tool calls...
2026-04-09 17:20:36.937 [INFO ] client.py:753 - Calling write_file with args: {"path": "docker-compose.yml", "content": "version: '3.8'\n\nservices:\n cadvisor:\n image: gcr.io/cadvisor/cadvisor:v0.49.1\n container_name: cadvisor\n ports:\n - "8080:8080"\n v
2026-04-09 17:20:36.960 [INFO ] client.py:814 - Tool execution completed, 1 tools executed
2026-04-09 17:20:36.960 [INFO ] grounding_agent.py:349 - Iteration 2 - Has tool calls: True, Tool results: 1, Content length: 0 chars
2026-04-09 17:20:36.961 [INFO ] grounding_agent.py:283 - Grounding Agent: Iteration 3/30
2026-04-09 17:20:36.961 [INFO ] client.py:166 - [SCHEMA_SANITIZE] Prepared 10 tools for LLM (from 10 total)
2026-04-09 17:20:36.962 [INFO ] utils.py:3995 -
LiteLLM completion() model= doubao-seed-2.0-pro; provider = openai
2026-04-09 17:20:40.628 [INFO ] client.py:703 - Executing 1 tool calls...
2026-04-09 17:20:40.629 [INFO ] client.py:753 - Calling run_shell with args: {"command": "mkdir -p prometheus grafana/provisioning/datasources grafana/provisioning/dashboards"}
2026-04-09 17:20:40.642 [INFO ] client.py:814 - Tool execution completed, 1 tools executed
2026-04-09 17:20:40.643 [INFO ] grounding_agent.py:349 - Iteration 3 - Has tool calls: True, Tool results: 1, Content length: 0 chars
2026-04-09 17:20:40.644 [INFO ] grounding_agent.py:283 - Grounding Agent: Iteration 4/30
2026-04-09 17:20:40.644 [INFO ] client.py:166 - [SCHEMA_SANITIZE] Prepared 10 tools for LLM (from 10 total)
2026-04-09 17:20:40.645 [INFO ] utils.py:3995 -
LiteLLM completion() model= doubao-seed-2.0-pro; provider = openai
2026-04-09 17:20:44.048 [INFO ] client.py:703 - Executing 1 tool calls...
2026-04-09 17:20:44.048 [INFO ] client.py:753 - Calling write_file with args: {"path": "prometheus/prometheus.yml", "content": "global:\n scrape_interval: 15s\n evaluation_interval: 15s\n\nscrape_configs:\n - job_name: 'cadvisor'\n static_configs:\n - targets: ['cadv
2026-04-09 17:20:44.072 [INFO ] client.py:814 - Tool execution completed, 1 tools executed
2026-04-09 17:20:44.072 [INFO ] grounding_agent.py:349 - Iteration 4 - Has tool calls: True, Tool results: 1, Content length: 0 chars
2026-04-09 17:20:44.072 [INFO ] grounding_agent.py:283 - Grounding Agent: Iteration 5/30
2026-04-09 17:20:44.073 [INFO ] client.py:166 - [SCHEMA_SANITIZE] Prepared 10 tools for LLM (from 10 total)
2026-04-09 17:20:44.073 [INFO ] utils.py:3995 -
LiteLLM completion() model= doubao-seed-2.0-pro; provider = openai
2026-04-09 17:20:47.640 [INFO ] client.py:703 - Executing 1 tool calls...
2026-04-09 17:20:47.640 [INFO ] client.py:753 - Calling write_file with args: {"path": "grafana/provisioning/datasources/prometheus.yml", "content": "apiVersion: 1\n\ndatasources:\n - name: Prometheus\n type: prometheus\n access: proxy\n url: http://prometheus:9090\n
2026-04-09 17:20:47.668 [INFO ] client.py:814 - Tool execution completed, 1 tools executed
2026-04-09 17:20:47.669 [INFO ] grounding_agent.py:349 - Iteration 5 - Has tool calls: True, Tool results: 1, Content length: 0 chars
2026-04-09 17:20:47.669 [INFO ] grounding_agent.py:283 - Grounding Agent: Iteration 6/30
2026-04-09 17:20:47.669 [INFO ] client.py:166 - [SCHEMA_SANITIZE] Prepared 10 tools for LLM (from 10 total)
2026-04-09 17:20:47.671 [INFO ] utils.py:3995 -
LiteLLM completion() model= doubao-seed-2.0-pro; provider = openai
2026-04-09 17:20:50.341 [INFO ] client.py:703 - Executing 1 tool calls...
2026-04-09 17:20:50.341 [INFO ] client.py:753 - Calling run_shell with args: {"command": "docker compose up -d"}
2026-04-09 17:21:05.669 [INFO ] client.py:814 - Tool execution completed, 1 tools executed
2026-04-09 17:21:05.669 [INFO ] grounding_agent.py:349 - Iteration 6 - Has tool calls: True, Tool results: 1, Content length: 0 chars
2026-04-09 17:21:05.670 [INFO ] grounding_agent.py:283 - Grounding Agent: Iteration 7/30
2026-04-09 17:21:05.670 [INFO ] client.py:166 - [SCHEMA_SANITIZE] Prepared 10 tools for LLM (from 10 total)
2026-04-09 17:21:05.671 [INFO ] utils.py:3995 -
LiteLLM completion() model= doubao-seed-2.0-pro; provider = openai
2026-04-09 17:21:13.937 [INFO ] client.py:703 - Executing 1 tool calls...
2026-04-09 17:21:13.937 [INFO ] client.py:753 - Calling shell_agent with args: {"task": "Fix and start the Docker Compose monitoring stack (cadvisor, prometheus, grafana) defined in docker-compose.yml in current directory. Verify all containers are running, and confirm the Grafa
2026-04-09 17:21:13.945 [INFO ] utils.py:3995 -
LiteLLM completion() model= doubao-seed-2.0-pro; provider = openai
2026-04-09 17:21:16.055 [INFO ] utils.py:3995 -
LiteLLM completion() model= doubao-seed-2.0-pro; provider = openai

While attempting to use OpenSpace to manage a Docker containers stack (including cadvisor, prometheus, and grafana), the agent remained in a "running" state for over 40 minutes without any error output or progress updates.
The shell_agent appears to lack a robust timeout mechanism or real-time output forwarding. In my case, the execution was likely blocked by a pending sudo password prompt or a stalled docker pull process in the background. Because OpenSpace waits for the subprocess to exit before reporting, the UI/CLI stays completely silent, leading to a "deadlock" user experience.