Chronicle Workflow

Chronicle Workflow

This skill guides you through the complete Chronicle workflow for tracking and documenting development work. Primarily uses CLI commands for portability, with optional MCP tools for faster programmatic access.

Auto-Activation

This skill auto-activates! (Milestone #13)

Prompts like "start a new session" or "is this tracked?" automatically trigger this skill. No manual loading needed!

Trigger patterns: start session, is this tracked, chronicle workflow, setup See: docs/HOOKS.md for full details

When to Use This Skill

Use this skill when:

• User is starting a new development session
• Setting up Chronicle for the first time
• Want to ensure all work is being tracked
• Need guidance on Chronicle best practices
• Building a comprehensive development knowledge base

Complete Workflow

1. Session Start (If Not Already Tracking)

Check if in Chronicle session:

• Current session is likely NOT being tracked unless started with chronicle start claude
• To track current work, user must exit and restart with Chronicle

Guide user:

# Exit current session
exit

# Start new Chronicle-tracked session
chronicle start claude

# Or for other tools:
chronicle start gemini

Important: Chronicle sessions must be started explicitly - they don't auto-track.

2. During Development

Best Practices:

• Work naturally - Chronicle captures everything automatically
• Make frequent git commits with descriptive messages
• Chronicle links commits to sessions automatically (±30 min window)

Available Commands (don't interrupt work to run these):

chronicle sessions              # List recent sessions
chronicle show today           # See today's commits
chronicle timeline today       # Combined view

3. Session End

Automatic on Exit:

• Full transcript captured to ~/.ai-session/sessions/session_N.log
• Metadata saved (duration, timestamp, repo)
• Session record created in database

What happens:

📊 Session #{id} complete! Duration: {minutes} minutes
💾 Full transcript saved
✨ Use 'chronicle session {id}' to view

4. Post-Session Documentation

Generate Summary (Automatic on first view):

# View session and auto-generate summary
chronicle session {id}

# For very large sessions (>50K lines)
chronicle summarize-chunked {id}

Document to Obsidian: Use the chronicle-session-documenter skill or manually:

"Document session {id} to my Obsidian vault"

5. Retrieval & Context

Find Past Work: Use the chronicle-context-retriever skill:

"How did I implement authentication last time?"
"What was the blocker with database migrations?"
"Show me all work on the API refactor"

Browse Sessions (CLI):

# Filter by repo
chronicle sessions --repo /path/to/project

# View timeline
chronicle timeline week

# Search sessions
chronicle search "authentication" --limit 5

Browse Sessions (MCP - if available):

# Faster programmatic access
sessions = mcp__chronicle__get_sessions(repo_path="/path/to/project", limit=20)
results = mcp__chronicle__search_sessions(query="authentication", limit=5)

Workflow Patterns

Pattern 1: Quick Feature Work

1. chronicle start claude
2. [Make changes, commit]
3. exit
4. [Summary auto-generates on next view]

Pattern 2: Multi-Session Project

1. chronicle start claude (Day 1)
2. [Work, commit]
3. exit
4. Document to Obsidian
5. chronicle start claude (Day 2)
6. "What did I do yesterday?" (retrieves context)
7. [Continue work]

Pattern 3: Research & Context

1. "Show me all sessions about X" (context retriever)
2. Review past approaches and decisions
3. chronicle start claude
4. [Implement with informed approach]

Multi-Project Tracking

Automatic Detection:

• Sessions detect working directory and git repo
• Stored in database for filtering

Filter by Project:

chronicle sessions --repo /path/to/project
chronicle timeline today --repo /path/to/project
chronicle summarize today --repo /path/to/project

Key Features to Highlight

Chunked Summarization

• Handles sessions of unlimited size
• Tested with 83,000+ line sessions
• Uses rolling summaries (10,000 lines/chunk)
• No rate limit issues

Obsidian Integration

• Full MCP server integration
• Create structured session notes
• Search past work
• Build knowledge graph
• Wikilinks between sessions/commits/repos

Local-First

• Everything stored locally (~/.ai-session/)
• SQLite database
• No data leaves your machine (except AI summarization)

Common Questions

Q: "Is this session being tracked?" A: Only if started with chronicle start claude. Must be explicit.

Q: "Can I track a session retroactively?" A: No - must start with Chronicle from the beginning.

Q: "How do I see what I did yesterday?" A: Use chronicle timeline yesterday or search Obsidian vault

Q: "How do I link commits to sessions?" A: Automatic! Commits within ±30 minutes of session are linked.

Database Location

All data stored at:

~/.ai-session/
├── sessions.db              # Main database
├── sessions/
│   ├── session_N.log       # Transcripts
│   └── session_N.meta      # Metadata
└── config.yaml             # Configuration

Pro Tips

1. Start Chronicle early - Can't retroactively track
2. Commit frequently - Better session-commit linking
3. Document important sessions - Build knowledge base in Obsidian
4. Search before building - Use context retriever to avoid repeating work
5. Use tags in Obsidian - Makes finding past work easier
6. Review summaries - Helps remember key decisions

Integration with Other Skills

• Use chronicle-session-documenter after completing sessions
• Use chronicle-context-retriever when starting related work
• Combine with git workflows for comprehensive tracking