skills.rest
nvie
nvieCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

self-review

Analyze Claude Code session history to surface repeated user corrections and preferences for CLAUDE.md.

3|Updated Apr 14, 2026
Search Tags:#workflow automation#self-improvement#claude code#claude.md#session analysis#user preferences#jsonl parsing
One-click install
npx skills add https://github.com/nvie/skills --skill self-review-nvie
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: self-review
Source: https://github.com/nvie/skills/tree/main/skills/self-review
Command: npx skills add https://github.com/nvie/skills --skill self-review-nvie

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Users often repeat the same corrections and preferences across Claude Code sessions without realizing these patterns should be persisted. This Skill mines past session history to find recurring feedback and turns it into durable CLAUDE.md rules or memory entries. ## Core Features & Use Cases - Session Mining: Scans JSONL session files under ~/.claude/projects/ and ~/.claude/sessions/, skipping subagent sessions and meta messages. - Pattern Detection: Identifies corrections, repeated preferences, contextual guidance, and frustration signals, prioritizing patterns that recur across multiple projects. - Guided Persistence: Classifies each finding as a global CLAUDE.md, project CLAUDE.md, or memory candidate, then walks through each item interactively for approval before applying changes. - Use Case: After weeks of using Claude Code, run a self-review to discover you have corrected the same git commit style five times across three projects, then add that rule to your global CLAUDE.md once instead of repeating it forever. ## Quick Start Review my past Claude Code sessions and find repeated corrections or preferences that should be added to my CLAUDE.md.

Frequently Asked Questions about self-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I find repeated corrections in my Claude Code sessions?▼

Run a self-review pass that scans JSONL session files under ~/.claude/projects/ and ~/.claude/sessions/. It extracts user messages, filters for correction and preference signals, and ranks patterns by frequency across sessions.

How to turn recurring Claude Code feedback into CLAUDE.md rules?▼

The review classifies each recurring pattern as a global CLAUDE.md, project CLAUDE.md, or memory candidate. It then walks through findings one by one, proposing exact text and applying changes only after your explicit approval.

Does session analysis include subagent or system messages?▼

No. Sessions under */subagents/* are skipped because they are agent-spawned, and messages with isMeta: true are excluded since they are system-injected rather than typed by the user.

What signals indicate a preference should be persisted?▼

Strong signals include corrections like "don't" or "I said", guidance like "always" or "from now on", and the same correction appearing in two or more unrelated projects, which indicates a global CLAUDE.md candidate.

Can session review expose sensitive information?▼

Session data may contain sensitive content, so the review only surfaces the specific quotes needed for each finding. It avoids dumping raw session content and flags ambiguous cases rather than assuming intent.

Related Skills

Explore More Software Engineering Skills→
ArthTech-AIArthTech-AI

continue

Parse Claude Code session transcripts to produce continuity briefings and next-step summaries.

Official
Advanced
icyrainzicyrainz

learn

Extract user preferences and lessons from conversations into CLAUDE.md and SKILL.md files.

Community
Advanced
Datus-aiDatus-ai
1.5k

memory-organization

Audit Datus persistent data stores and generate a remediation plan.

Official
Advanced
kevinnftkevinnft
13

context-engineering

Load project-wide context rules, specs, and source files for agent sessions.

Community
Advanced