Engineering
changelog avatar

changelog

Automated runtime observability changelog for Claude Code development sessions, tracking file changes, test results, and git commits.

Introduction

The changelog skill serves as an advanced audit and observability tool for developers using Claude Code. By integrating directly into the development loop via PostToolUse hooks, it automatically captures a chronological stream of events as you work. This allows subagents, debuggers, and human developers to reconstruct the sequence of actions that led to a project's current state, effectively answering the question of how a specific project configuration was reached. The skill maintains a persistent JSONL (JSON Lines) log file, ensuring that event data is structured for both machine readability and human analysis. It acts as an essential companion to the auto-loop checkpoint system, providing a historical context stream that complements the static snapshot of a checkpoint. Targeted at software engineers and technical teams, it is particularly useful for debugging complex failures, performing post-mortems on development sessions, and providing context to multi-agent team workflows.

  • Automatically logs file creation and modification events using PostToolUse hooks for real-time observability.

  • Records bash-level activities including test execution outcomes (pass/fail) and git commit events.

  • Implements automated log rotation logic, keeping the current log file under 500 lines to prevent performance degradation or excessive disk usage.

  • Facilitates session recovery by enabling agents to query historical logs to understand previous project interactions and environment states.

  • Supports portable path management through the use of CLAUDE_PROJECT_DIR environment variables.

  • Ensure the skill is active in your project environment to enable automatic hook triggering on Write, Edit, and Bash tool usage.

  • Review the generated .director-mode/changelog.jsonl file for a complete audit trail of the current session.

  • If hooks are bypassed or fail, developers can trigger event logging manually via auto-loop prompts to maintain consistency in the audit stream.

  • The system enforces a single auto-loop session per project; use the --resume or --force flags to manage session conflicts or re-initialize logs.

  • Archive files are stored with timestamped filenames, allowing for long-term historical analysis of project evolution over time.

Repository Stats

Stars
66
Forks
8
Open Issues
3
Language
Shell
Default Branch
main
Sync Status
Idle
Last Synced
May 3, 2026, 03:36 PM
View on GitHub