orchestration-logger

majiayu000's avatarfrom majiayu000

Unified logging system for AI orchestration workflows with current state metadata and execution logs in a single file

⭐5stars🔀1forks📁View on GitHub🕐Updated Jan 11, 2026
[Agent Observability]

When & Why to Use This Skill

The Orchestration Logger is a specialized Claude skill designed to provide a unified logging system for AI orchestration workflows. It enables seamless tracking of execution states, metadata, and subagent interactions in a single JSON file, offering a transparent audit trail and real-time progress monitoring for complex agentic tasks without requiring external scripts.

Use Cases

  • Debugging Multi-Agent Workflows: Trace every request and response between subagents to identify logic errors or bottlenecks in complex AI orchestration loops.
  • Audit and Compliance: Maintain a detailed, timestamped record of AI-driven decisions and execution history for quality assurance and regulatory reporting.
  • Real-Time Progress Monitoring: Track the status of long-running tasks, including completion percentages, current subagent assignments, and task success/failure rates through structured metadata.
  • State Management for AI Agents: Provide a persistent 'source of truth' for the current state of an orchestration process, allowing agents to resume or pivot based on previous execution logs.
nameorchestration-logger
descriptionUnified logging system for AI orchestration workflows with current state metadata and execution logs in a single file

Orchestration Logging Skill

Unified logging for orchestration workflows with automatic state tracking and audit trails. You will create and append orchestration logs to the logfile by directly updating the json file. No code or scripts required.

Instructions

  1. Orchestration logs should be generated under .claude/logs with the filename .claude/logs/orchestration-{app_name}-{yyyymmdd-HH24MMss}.json containing:
  • Metadata: Orchestration progress, completed tasks, current status
  • Execution logs: Complete logging of all subagent requests and responses
  • Timestamp should be in SGT

Usage Patterns

  1. Initialize- Start by creating a new logfile with the json structure and metadata
  2. Log Request- Log the request under execution_logs with type:request. Update metadata
  3. Log Response- Log the response under execution_logs with type:response. Update metadata

Example Log

Follow this sample log for guidance 

{
  "orchestration_id": "{orchestration_id}",
  "log_file": "orchestration-{orchestration_id}.json",
  "metadata": {
    "status": "{running|completed|halted|error}", # halted : if waiting for user input
    "app_name": "hello-world",
    "confirmation_mode": "interactive",
    "started_at": "{timestamp in SGT}",
    "implementation_plan": "plans/us-hello-impl-plan.md",
    "user_story_id": "US-HELLO",
    "current_task": "1.2",
    "current_subagent": "quality-manager",
    "completed_tasks": [1.1],
    "failed_tasks": [],
    "loop_attempts": {},
    "total_tasks": 82,
    "completion_percentage": 1.05,
    "last_completed_subtask_at": null,
    "last_completed_task_id": null,
    "last_updated": "{timestamp in SGT}",
    "final_report": null
  },
  "execution_logs": [
    {
      "task": "1.2",
      "subagent": "python-developer",
      "type":"request",        
      "content":"Implement task 1.2 | Fix the following issues : (a), (b), (c)..",
      "timestamp":"{timestamp in SGT}",
      "implementation_plan": "plans/us-hello-impl-plan.md",
      "user_story": "us-hello-world.md",
    },
    {
      "task": "1.2",
      "subagent": "python-developer",
      "type":"response",        
      "content":"I have implemented task 1.2 with the following changes (a), (b), (c)..",
      "timestamp":"{timestamp in SGT}",
    },
    {
      "task": "1.2",
      "subagent": "quality-manager",
      "type":"request",        
      "content":"Validate the implementation for task 1.2 against the acceptance criteria",
      "timestamp":"{timestamp in SGT}",
      "implementation_plan": "plans/us-hello-impl-plan.md",
      "user_story": "us-hello-world.md",
    },

  ]
}