replay-learnings

Searches correction history and past learnings to surface relevant patterns before starting a task.

2.8k|277|Updated Feb 1, 2026
One-click install
npx skills add https://github.com/rohitg00/pro-workflow --skill replay-learnings
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: replay-learnings
Source: https://github.com/rohitg00/pro-workflow/tree/main/skills/replay-learnings
Command: npx skills add https://github.com/rohitg00/pro-workflow --skill replay-learnings

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Developers repeat the same mistakes across sessions because past corrections and lessons learned are scattered or forgotten. This Skill searches your stored learnings and session history before you start work, so prior mistakes and patterns inform the current task.

Core Features & Use Cases

  • Keyword-Based Learning Search: Extracts keywords from your task description and greps learning files like LEARNED.md, learning-log.md, and CLAUDE.md for matching patterns.
  • Relevance-Ranked Briefing: Surfaces top learnings ranked by relevance with original mistake context and application counts.
  • Session History Analysis: Reviews past sessions for similar work and flags high correction-rate sessions requiring extra care.
  • Use Case: Before starting an auth refactor, run a replay to recall that external APIs must be mocked in auth tests and that auth middleware lives in src/middleware/, avoiding repeated corrections.

Quick Start

Ask the assistant to replay what it knows about your upcoming task, for example: "What do I know about auth refactoring before I start?"

Frequently Asked Questions about replay-learnings

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

FAQPage Schema
How do I recall past coding mistakes before starting a new task?

Run a replay of your learnings by describing the upcoming task. The Skill extracts keywords, searches your learning files for matching patterns, and returns a ranked briefing of relevant past corrections with their original mistake context.

What files does the learning search look through?

It searches .claude/LEARNED.md, .claude/learning-log.md, and entries marked with [LEARN] in CLAUDE.md using case-insensitive grep on keywords extracted from your task description.

How are past learnings ranked in the replay output?

Learnings are ranked by relevance to the current task, not recency. Each entry includes the original mistake context and how many times the learning has been applied, so the most actionable patterns surface first.

What happens if no learnings match my task?

The Skill explicitly reports that no learnings were found rather than forcing irrelevant results. It then suggests starting with a scout agent to explore the codebase first.

Can it warn me about tasks where I historically made many mistakes?

Yes. It reviews session history for similar work and flags sessions with high correction rates, so you know to review patterns carefully before starting tasks that previously caused repeated corrections.