How does it handle logging without overwriting files?

The skill provides specific Bash patterns using append operators (>>) to ensure that logs are added sequentially throughout the workflow session.

What are DECISION logs in this context?

DECISION logs record critical branching points, such as AI confidence scores and user responses, to provide a clear audit trail of why a specific path was chosen.

Can I use this for multi-agent coordination?

Absolutely. It includes specific event types for 'Master Coordinators' to pass logging context down to 'Phase Agents,' ensuring unified logging across different agents.

Does this skill support machine-readable formats?

Yes, it specifically defines a JSONL (JSON Lines) format that is ideal for ingestion by log analysis tools and programmatic monitoring.

What is the primary purpose of the workflow-logging skill?

It provides a standardized schema and implementation method for agents to record their progress, making complex AI workflows observable and easier to debug.

Workflow Logging Specification

Name: Workflow Logging Specification
Author: penkzhou

bypenkzhou

•

Analytics & Monitoring

Standardizes Claude Code workflow monitoring through structured JSONL and human-readable text logging patterns.

About

The workflow-logging skill provides a comprehensive framework for tracking and auditing complex AI-driven development tasks. It defines a rigorous specification for recording session events, agent interactions, tool usage, and critical decision points. By supporting dual formats—structured JSONL for programmatic analysis and a clean text format for human review—it ensures full transparency into how Claude processes tasks. This skill is essential for developers building multi-agent systems who need reliable observability and debugging capabilities for automated workflows.

Key Features

1 GitHub stars
Standardized logging levels including INFO, DEBUG, WARN, ERROR, and DECISION
Implementation patterns for non-destructive log appending via Bash
Dual-format support for structured JSONL and human-readable text logs
Comprehensive metadata capturing for environment, confidence scores, and user interactions
Granular event tracking for sessions, phases, agents, and tool usage

Use Cases

Debugging agent failures by reviewing granular step-by-step tool execution history
Building observability dashboards for automated AI development pipelines
Auditing complex bug-fix workflows to understand AI decision-making logic

About

Key Features

1 GitHub stars
Standardized logging levels including INFO, DEBUG, WARN, ERROR, and DECISION
Implementation patterns for non-destructive log appending via Bash
Dual-format support for structured JSONL and human-readable text logs
Comprehensive metadata capturing for environment, confidence scores, and user interactions
Granular event tracking for sessions, phases, agents, and tool usage

Use Cases

Debugging agent failures by reviewing granular step-by-step tool execution history
Building observability dashboards for automated AI development pipelines
Auditing complex bug-fix workflows to understand AI decision-making logic