Does it store my session data in the cloud?

No, all indexing and searching are performed locally on your machine using DuckDB and your local filesystem.

Do I need to manually index my sessions every time?

The index is automatically initialized on your first 'list' command, but you should run the 'index' command or pass '--rebuild' periodically to capture the most recent sessions.

How do I find a specific experiment I mentioned last week?

You can use the 'search' command with a keyword (e.g., 'session_query.py search "experiment"') to scan all raw transcripts for that specific topic.

How does Session Historian access my session data?

It reads local JSONL transcript files stored in your standard Claude project directory (~/.claude/projects/) and indexes them into a local DuckDB database for fast querying.

Can it summarize what a sub-agent did in a past session?

Yes, it tracks sub-agent IDs, their specific tasks, and whether they succeeded, failed, or were blocked, presenting this in a structured markdown summary.

Session Historian

Name: Session Historian
Author: Jamie-BitFlight

byJamie-BitFlight

•

生産性とワークフロー

Recalls and analyzes past Claude Code sessions by indexing local transcripts and generating structured activity summaries.

Session Historian bridges the context gap between separate Claude Code interactions by providing a searchable index of past conversations. It allows developers to quickly retrieve forgotten decisions, track sub-agent outcomes, and audit tool performance from previous sessions. By indexing raw JSONL transcripts into a local DuckDB database, it enables high-speed searches and provides verbatim message recovery, ensuring that project continuity is maintained even when the AI's short-term memory is cleared. This is an essential utility for long-running projects where maintaining a clear audit trail of experiments and decisions is critical.

主な機能

01Full-text search across all prior Claude session transcripts via DuckDB

02Verbatim retrieval of user messages to reconstruct original context and intent

03Diagnostic tools to detect tool errors and recursive 'stuck' loops

0428 GitHub stars

05Structured session summaries including AI actions and sub-agent outcomes

06Frustration detection to identify and audit past points of friction

ユースケース

01Reviewing sub-agent task outcomes to understand how complex workflows were completed

02Recovering architectural decisions or experimental results from a session several days ago

03Troubleshooting recurring tool failures by analyzing success/failure breakdowns across sessions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jamie-bitflight/claude_skills session-historian

For use in Claude.ai and ChatGPT

Download Skill