Can I control how much history is analyzed?

Yes, you can use arguments like --since-days and --max-pairs to limit the scan range, ensuring the skill only focuses on recent or relevant conversation history.

Does it automatically overwrite my existing documentation?

No. The skill follows a strict multi-phase process where it presents a summary of all proposed changes first. It requires your explicit approval via an interactive prompt before any files are modified.

Where does the extracted knowledge get stored?

Knowledge is routed based on scope: universal rules go to CLAUDE.md, domain-specific patterns go to the docs folder, and path-scoped conventions are placed in .claude/rules/.

How does the skill identify specific learnings?

The skill uses an extraction script to scan past transcripts for explicit user corrections, pattern-based discoveries, and mandates that indicate a gap between the AI's behavior and the project's actual requirements.

Session History Miner

Name: Session History Miner
Author: propstreet

bypropstreet

•

学習とドキュメンテーション

Extracts and synthesizes persistent knowledge from past Claude Code sessions into actionable project documentation.

The Session History Miner automates the process of turning fleeting conversational corrections into permanent project rules by scanning Claude Code history. It analyzes past transcripts to identify recurring patterns, user corrections, and domain-specific discoveries, then intelligently proposes updates to files like CLAUDE.md or dedicated documentation folders. This skill helps developers maintain a 'living' knowledge base, preventing recurring mistakes and ensuring that the tribal knowledge gained during AI coding sessions is codified for future use.

主な機能

01Automated transcript scanning for user corrections and recurring coding patterns

022 GitHub stars

03Pre-update documentation audit to prevent duplication and resolve contradictions

04Multi-level routing to CLAUDE.md, domain-specific docs, or path-scoped rules

05Intelligent synthesis that filters session-specific noise from universal rules

06Interactive proposal summary with manual approval before any file modifications

ユースケース

01Maintaining a high-quality developer handbook that evolves automatically based on real coding history

02Codifying recurring AI corrections into CLAUDE.md to prevent the same mistakes in future sessions

03Onboarding new team members by formalizing undocumented project quirks discovered during development

主な機能

01Automated transcript scanning for user corrections and recurring coding patterns

022 GitHub stars

03Pre-update documentation audit to prevent duplication and resolve contradictions

04Multi-level routing to CLAUDE.md, domain-specific docs, or path-scoped rules

05Intelligent synthesis that filters session-specific noise from universal rules

06Interactive proposal summary with manual approval before any file modifications

ユースケース

01Maintaining a high-quality developer handbook that evolves automatically based on real coding history

02Codifying recurring AI corrections into CLAUDE.md to prevent the same mistakes in future sessions

03Onboarding new team members by formalizing undocumented project quirks discovered during development